# Maple full guide

The public marketing site lives on https://www.trymaple.ai. The application and login surface remain on https://trymaple.ai until the app is ready to move to https://app.trymaple.ai.

## Current public site scope
- Live pages: https://www.trymaple.ai/, https://www.trymaple.ai/about, https://www.trymaple.ai/agent, https://www.trymaple.ai/blog, https://www.trymaple.ai/blog/feed, https://www.trymaple.ai/blog/authors, https://www.trymaple.ai/blog/authors/mark-suman, https://www.trymaple.ai/blog/authors/anthony-ronning, https://www.trymaple.ai/blog/authors/michael-atwood, https://www.trymaple.ai/brand, https://www.trymaple.ai/careers, https://www.trymaple.ai/downloads, https://www.trymaple.ai/learn/models, https://www.trymaple.ai/pricing, https://www.trymaple.ai/privacy, https://www.trymaple.ai/proof, https://www.trymaple.ai/proxy-api, https://www.trymaple.ai/research, https://www.trymaple.ai/status, https://www.trymaple.ai/teams, https://www.trymaple.ai/terms
- All other internal marketing routes are intentionally unresolved and currently return 404 until they receive a real design pass.

## Homepage summary
- Hero: Personal AI for your real life
- Value prop: The personal AI you can actually trust with your home life, your work, and what you'd never tell Big AI.

## Homepage sections
- Header / top navigation
- Hero
- Product spotlight tiles
- Privacy-first architecture
- Comparison table
- Final CTA
- Footer

## Product spotlights
- Research: The secure workspace for life's harder questions. Long conversations, document analysis, image analysis, kept private to you.
- Agent: Maple remembers everything, making it a useful companion. The more you tell it, the more it can help you.

## Agent page summary
- Title: Maple Agent Waitlist
- Value prop: Maple Agent is the personal AI for your real life. It remembers what matters and keeps every detail between the two of you. Join the waitlist for early access.
- Primary action: join the Maple Agent waitlist with an email address.
- Form handling: static Formspree submission for early-access reservations.

## Careers page summary
- Title: Careers at Maple
- Hero: Make open models easy to use, access privately, and verify.
- Mission: We are bringing powerful open models to more people through approachable products and privacy protections you can check for yourself.
- The work: Capable open-weight models should be available to more than people who can run their own infrastructure. We are building the product and the verifiable privacy architecture that make them practical for everyday life and work.
- Bring open models to more people: Make capable open-weight models accessible without asking people to operate their own infrastructure.
- Make powerful AI feel simple: Turn model choice, documents, and complex workflows into an experience people can use every day.
- Make trust verifiable: Protect private work in secure enclaves, with open-source code and live attestation people can check for themselves.
- Open roles: We do not have any open positions at the moment. When we are hiring, you will find the roles and application details here.

## Research page summary
- Title: Maple Research
- Value prop: A private workspace for the research and writing in your life. Long-form chat, document analysis, and web search, all in one place.
- Plans and apps: https://www.trymaple.ai/pricing and https://www.trymaple.ai/downloads. Section ids #pricing and #download stay on this page.

## Maple Proxy page summary
- Title: Maple Proxy | Private AI for the tools you already use
- Value prop: Maple Proxy is an OpenAI-compatible endpoint that runs on your own machine. It verifies Maple’s secure enclave and encrypts every request, so the clients you already use keep working.
- Primary action: Get the desktop app at https://www.trymaple.ai/research#download.
- The private part, handled for you. Talking to Maple privately takes two extra steps: checking that the server is the secure enclave it claims to be, and encrypting what you send. Most tools don’t know how to do either. Maple Proxy does both, and gives your tools the endpoint they already expect.
- For tools on your own machine: Point a coding agent, an editor, or a script at the proxy. If it works with OpenAI’s chat, embeddings, or models endpoints, it can talk to Maple.
- For developers shipping apps: Run the proxy next to your service and keep the OpenAI client library you already build with.
- Open source: The proxy’s code is public. Read how it verifies the enclave and encrypts your requests, or build it yourself.
- How a request moves: From your tool to Maple and back, in four steps.
- Your tool calls the proxy: It sends a normal OpenAI-style request, with your Maple API key, to the proxy on your machine.
- The proxy checks the enclave: Before anything is sent, it verifies Maple’s enclave attestation and agrees on encryption keys with it.
- The request leaves encrypted: Your prompt and your API key are sealed on your machine, then sent to the enclave.
- The answer comes back: The proxy decrypts the response and hands it to your tool, streamed or all at once.
- Encryption starts at the proxy. The hop from your tool to the proxy is ordinary local traffic, which is why the proxy runs on a machine you control. See how Maple’s enclaves work at https://www.trymaple.ai/proof.
- Start it from the app, or run it yourself. Both run the same open-source proxy. Pick the one that matches where your tools live.
- Local Proxy: Turn on Local Proxy in the Maple desktop app’s settings. It starts the endpoint on your computer and can create an API key for it.
- Docker: Run the Maple Proxy container beside your own service, on your laptop or on a server you manage.
- Call Maple directly with the SDK: Building in JavaScript? @mapleai/sdk does the same verification and encryption inside your app, so there is nothing extra to run.
- Pricing: Plans, API access, and per-model rates. See pricing at https://www.trymaple.ai/research#pricing.
- Setup guide: Step-by-step setup for the desktop app and Docker. Read the guide at https://www.trymaple.ai/docs/proxy.
- Source code: The proxy and the SDKs, in Maple’s public repository. View on GitHub at https://github.com/MaplePrivacyLabs/Maple/tree/master/proxy.
- Final action: Get started at https://trymaple.ai.

## Developer docs
- Static developer documentation on the marketing host. The full text of every page follows. Each page is also served as Markdown at its URL plus .md.
- Maple Proxy: Setup, authentication, endpoints and configuration for Maple Proxy, the OpenAI-compatible server that verifies Maple’s secure enclave and encrypts each request.

### Maple docs

> Set up Maple Proxy, connect your tools, and choose models by their context window, thinking controls, image support and other practical capabilities.

Source: <https://www.trymaple.ai/docs> · Last updated: 2026-10-08

Use Maple from your own tools and applications. Start with the desktop app's Local Proxy, or run a dedicated proxy with Docker or the release binary.

#### Maple Proxy

- [Maple Proxy](https://www.trymaple.ai/docs/proxy): Start the proxy and connect an OpenAI-compatible tool.
- [Models and capabilities](https://www.trymaple.ai/docs/proxy/models): Model IDs, context windows, thinking levels, images and tool use.

For plans and API rates, see [pricing](https://www.trymaple.ai/pricing). Get the apps from [downloads](https://www.trymaple.ai/downloads).

### Get started with Maple Proxy

> Connect your tools to Maple with the desktop app's local proxy. Set the base URL, choose a model, and send a first request, or run Docker or a binary yourself.

Source: <https://www.trymaple.ai/docs/proxy> · Last updated: 2026-10-08

Maple Proxy gives tools that support OpenAI Chat Completions a local endpoint for Maple's private models. It verifies Maple's secure enclave and encrypts traffic from the proxy to Maple.

API access requires Pro, Max or Team; API usage is billed separately. See [pricing](https://www.trymaple.ai/pricing).

#### Desktop app

The [Maple desktop app](https://www.trymaple.ai/downloads) includes the proxy for macOS, Windows and Linux.

1. [Download Maple](https://www.trymaple.ai/downloads) and sign in.
2. Open **Settings**, then **API & credits** under **Developer**.
3. Open the **Local Proxy** tab and select **Start proxy**.

Maple creates and saves a proxy key if you have not set one. Copy the base URL from this screen; the default is `http://127.0.0.1:8080/v1`. Keep Maple open while using connected tools.

With the default CORS setting off, this request uses the saved key because it omits `Authorization`:

```bash
curl -N http://127.0.0.1:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-oss-120b",
    "messages": [{"role": "user", "content": "Write a haiku about privacy"}],
    "stream": true
  }'
```

The first request can take longer while the proxy verifies the enclave. For problems starting it, see [troubleshooting](https://www.trymaple.ai/docs/proxy/troubleshooting).

##### Connect an application

Choose an **OpenAI-compatible / Chat Completions** provider and enter:

| Setting  | Value                                                                                                                                  |
| -------- | -------------------------------------------------------------------------------------------------------------------------------------- |
| Base URL | `http://127.0.0.1:8080/v1` (or the URL shown in Maple)                                                                                 |
| API key  | A **real Maple API key** if the tool requires one; [create a key](https://www.trymaple.ai/docs/proxy/authentication#create-an-api-key) |
| Model    | An exact ID from [models and capabilities](https://www.trymaple.ai/docs/proxy/models)                                                  |

Some applications expect the URL without `/v1`; check their instructions. A placeholder key overrides the saved key and fails authentication.

Follow the [OpenCode](https://www.trymaple.ai/docs/proxy/guides/opencode) or [Goose](https://www.trymaple.ai/docs/proxy/guides/goose) guide, or use the [API reference](https://www.trymaple.ai/docs/proxy/api) for your own application.

##### Desktop settings

Under **Advanced settings**, change the port, enable **Auto-start when Maple launches**, or select **Enable CORS for browser clients**. With CORS on, every client must send its own key. See [authentication](https://www.trymaple.ai/docs/proxy/authentication#browser-clients-and-cors).

> **Keep the proxy local**
>
> Traffic between your application and the proxy is plaintext. Keep the host on `127.0.0.1` for tools on your computer. If that traffic leaves your machine, protect it with TLS and a trusted network boundary.

#### Docker or binary

Run a dedicated proxy beside your service or on a machine without the desktop app. See [Docker and binary setup](https://www.trymaple.ai/docs/proxy/configuration).

### Run with Docker or a binary

> Run Maple Proxy independently with Docker or a release binary. Verify downloads, connect with your API key, and configure ports, browser access and timeouts.

Source: <https://www.trymaple.ai/docs/proxy/configuration> · Last updated: 2026-10-08

Use a dedicated proxy when your tools run on a server or without the [desktop app](https://www.trymaple.ai/docs/proxy#desktop-app). Keep it on the same machine as your application where possible: the application-to-proxy connection is plaintext.

#### Docker

The image supports AMD64 and ARM64. Publish its port only on loopback:

```bash
docker run --rm -p 127.0.0.1:8080:8080 \
  ghcr.io/mapleprivacylabs/maple-proxy:latest
```

The image enables CORS, so **each request must send its own API key**. It does not use `MAPLE_API_KEY` while CORS is on. It includes a local health check and connects to Maple's production service.

For a persistent deployment, pin the image tag or digest you have tested. Only for a private deployment where every caller is trusted, add `-e MAPLE_ENABLE_CORS=false -e MAPLE_API_KEY` after exporting your key. See [authentication](https://www.trymaple.ai/docs/proxy/authentication).

#### Binary

Download the archive for your platform and the [checksum file](https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-release-final.sha256). These links use Maple release **v3.4.3** (proxy **0.4.1**).

| Platform            | Download                                                                                                      |
| ------------------- | ------------------------------------------------------------------------------------------------------------- |
| Linux x86_64        | [tar.gz](https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-linux-x86_64.tar.gz)  |
| Linux ARM64         | [tar.gz](https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-linux-aarch64.tar.gz) |
| macOS Apple silicon | [tar.gz](https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-macos-aarch64.tar.gz) |
| Windows x86_64      | [zip](https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-windows-x86_64.zip)      |

Verify and extract the archive before running it. Linux x86_64:

```bash
curl -LO https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-linux-x86_64.tar.gz
curl -LO https://github.com/MaplePrivacyLabs/Maple/releases/download/v3.4.3/maple-proxy-release-final.sha256
sha256sum --check --ignore-missing maple-proxy-release-final.sha256
tar -xzf maple-proxy-linux-x86_64.tar.gz
export MAPLE_API_KEY='your-real-maple-api-key'
./maple-proxy
```

On macOS, verify with `shasum -a 256 --check --ignore-missing maple-proxy-release-final.sha256`. This binary is not notarized; after verifying and extracting it, remove the download quarantine with `xattr -d com.apple.quarantine ./maple-proxy` if macOS blocks it. On Windows, compare `Get-FileHash <archive> -Algorithm SHA256` with the checksum file, extract the ZIP, set `$env:MAPLE_API_KEY`, and run `./maple-proxy.exe`.

The binary defaults to `127.0.0.1:8080`, with CORS off. It can use the exported key for requests without an Authorization header.

#### Connect your client

[Create a Maple key](https://www.trymaple.ai/docs/proxy/authentication#create-an-api-key) and export it in the shell making requests. Starting the desktop app or a container does not set this shell variable:

```bash
export MAPLE_API_KEY='your-real-maple-api-key'
curl http://127.0.0.1:8080/v1/models \
  -H "Authorization: Bearer $MAPLE_API_KEY"
```

This checks the connection and credentials, and returns the API catalog; it does not prove model access or successful inference. Use `http://127.0.0.1:8080/v1` as the client base URL, your real key, and a [model suited to your task](https://www.trymaple.ai/docs/proxy/models). See the [API reference](https://www.trymaple.ai/docs/proxy/api).

#### Settings

Environment variables configure the standalone proxy; command-line flags take priority. Run `maple-proxy --help` for all options in your version.

| Setting                          | Default             | Use                                                                                                                                   |
| -------------------------------- | ------------------- | ------------------------------------------------------------------------------------------------------------------------------------- |
| `MAPLE_HOST`, `MAPLE_PORT`       | `127.0.0.1`, `8080` | Numeric listen address and port. Docker listens on `0.0.0.0` inside the container; control exposure with the published port.          |
| `MAPLE_ENABLE_CORS`              | Off; on in Docker   | Browser response access. Requires per-request keys when on.                                                                           |
| `MAPLE_API_KEY`                  | None                | Saved key for trusted callers when CORS is off.                                                                                       |
| `MAPLE_REQUEST_TIMEOUT_SECS`     | `300`               | Time before a response starts; also caps the full non-streaming response. Initial enclave verification has a separate timeout budget. |
| `MAPLE_STREAM_IDLE_TIMEOUT_SECS` | `300`               | Maximum gap between response chunks. Streaming has no total-duration cap.                                                             |
| `MAPLE_DEBUG`                    | `false`             | Enable request method/path debug logs when diagnosing a problem.                                                                      |

Both timeouts must be positive. A timeout after a response starts ends the body; do not assume retrying a request cannot run it twice.

Optionally keep `MAPLE_CACHE_NAMESPACE_ROOT` in your secrets manager to preserve cache continuity across restarts. Generate it once with `openssl rand -base64 32`; cache hits are not guaranteed. Keep it and API keys out of source control.

### Authentication and API keys

> Create and revoke Maple API keys, choose between a saved key and per-request authentication, and understand browser access for desktop and standalone proxies.

Source: <https://www.trymaple.ai/docs/proxy/authentication> · Last updated: 2026-10-08

Requests are billed to the account that owns the API key.

#### Create an API key

Open **Settings → API & credits → API Keys** in Maple. Create a key and copy it: the full key is shown only once. Give each tool its own key so you can revoke it separately. Delete lost or leaked keys here and create replacements; never commit keys to source control.

#### Which key is used?

- `Authorization: Bearer <your-key>` takes priority. Use a real Maple key; placeholders override the saved key.
- With CORS off, a request that **omits Authorization** can use the desktop's saved key or standalone `MAPLE_API_KEY`.
- With CORS on, every request needs its own key; the saved key is ignored. `/health` needs no key.

#### Browser clients and CORS

Enabling CORS permits any website origin. Desktop rejects browser-origin requests when CORS is off; standalone CORS-off is **not access control**.

Keep a saved key only within a trusted, private boundary: every caller can spend it. For shared or browser-facing proxies, omit the saved key and require per-request keys.

### Models and capabilities

> Choose a Maple model by its exact API ID, context window, thinking support, image input and tool calls, and check the current list through your local proxy.

Source: <https://www.trymaple.ai/docs/proxy/models> · Last updated: 2026-10-08

Use a model's exact ID in your request's `model` field. Choose one with the capabilities your application needs.

#### Chat models

Capabilities checked on 8 October 2026:

| Model / ID                                                        | Context tokens | Thinking | Image input | Tool calls |
| ----------------------------------------------------------------- | -------------- | -------- | ----------- | ---------- |
| GLM-5.3 Flash `glm-5-3-flash`                                     | 1,048,576      | Yes      | Yes         | Yes        |
| GLM 5.3 `glm-5-3`                                                 | 262,144        | Yes      | No          | Yes        |
| Kimi K3 `kimi-k3`                                                 | 262,144        | Yes      | Yes         | Yes        |
| DeepSeek V4.1 Flash `deepseek-v4-1-flash`                         | 1,048,576      | Yes      | Yes         | Yes        |
| OpenAI GPT-OSS 120B `gpt-oss-120b`                                | 131,072        | Yes      | No          | Yes        |
| Gemma 4 31B `gemma4-31b`                                          | 262,144        | Yes      | Yes         | Yes        |
| Llama 3.3 70B `llama3-3-70b`                                      | 131,072        | No       | No          | Yes        |
| OpenAI GPT-OSS Safeguard 120B (API only) `gpt-oss-safeguard-120b` | 131,072        | Yes      | No          | No         |

Context is the token window advertised by Maple, shared by the input, history, reasoning and generated answer. Leave room for the answer; the full window is not an input allowance or a guaranteed output length. Set `max_tokens` to bound the output of a chat completion.

Image input means the model accepts images through the chat API. Tool calls mean the model can request a function; your application executes it and returns the result.

See [pricing](https://www.trymaple.ai/pricing) for rates. A listed model can still fail a request because of your plan, quota or temporary availability.

#### Get the current list

With the proxy running:

```bash
curl http://127.0.0.1:8080/v1/models \
  -H "Authorization: Bearer $MAPLE_API_KEY"
```

Inspect `id`, `tasks`, `capabilities` and `context_window`. The list describes models; it does not guarantee that your account can use every entry. A `generate` task indicates a chat model; `embed` indicates an embedding model.

#### Thinking levels

Through the chat API, `reasoning_effort` controls a reasoning model's thinking budget. For `gpt-oss-120b`, use `low`, `medium` or `high`; the default is `medium`. Leave the field unset for the model's default. Other models accept different controls, so do not copy one model's levels to another. Higher effort generally takes more time and uses more output tokens.

#### Automatic selection

`auto:quick` and `auto:powerful` let Maple select a model for speed or deeper analysis. The selected model and its capabilities can change. Use an explicit model ID when your integration requires a particular context window or image support.

#### Embeddings

Use `nomic-embed-text` with `/v1/embeddings`. It produces vectors for search and retrieval, rather than chat responses. Other entries with `transcribe` or `speech` tasks are not chat models.

### API reference

> Call Maple Proxy with OpenAI-compatible requests. A short reference for chat, embeddings, model discovery, streaming and the limits that matter to clients.

Source: <https://www.trymaple.ai/docs/proxy/api> · Last updated: 2026-10-11

Set your tool's OpenAI base URL to `http://127.0.0.1:8080/v1`, use a real [Maple API key](https://www.trymaple.ai/docs/proxy/authentication), and choose a [model ID](https://www.trymaple.ai/docs/proxy/models). This also applies to custom OpenAI-compatible providers in LangChain, LlamaIndex and similar tools.

#### Supported endpoints

| Method | Path                   | Use                                                                                                            |
| ------ | ---------------------- | -------------------------------------------------------------------------------------------------------------- |
| `POST` | `/v1/chat/completions` | Chat, including streaming and model-supported tool/image inputs                                                |
| `POST` | `/v1/embeddings`       | Text embeddings with `nomic-embed-text`                                                                        |
| `POST` | `/v1/systemone`        | Typed decisions with calibrated probabilities; see [System One](https://www.trymaple.ai/docs/proxy/system-one) |
| `GET`  | `/v1/models`           | Model catalog and capability metadata                                                                          |
| `GET`  | `/health` (or `/`)     | Local process liveness and version; no key needed                                                              |

Chat, embeddings and models are available in the released proxy 0.4.1; `/v1/systemone` needs 0.4.2 or newer. Responses, Assistants, audio, image-generation and file-upload APIs are not supported. A healthy `/health` response does not check your key or the Maple connection.

#### Chat completions

Create a key in Maple, then set it in the shell where you run the request:

```bash
export MAPLE_API_KEY=your-maple-api-key
curl -N http://127.0.0.1:8080/v1/chat/completions \
  -H "Authorization: Bearer $MAPLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-oss-120b",
    "messages": [{"role": "user", "content": "Write a haiku about privacy"}],
    "reasoning_effort": "medium",
    "stream": true
  }'
```

Set `stream` to `false` for a single JSON response. Streaming uses Server-Sent Events. Models that expose reasoning return it in `delta.reasoning` while streaming or `message.reasoning` in a complete response; answer text is in `content`.

Supported thinking levels and image/tool inputs depend on the model. See [models and capabilities](https://www.trymaple.ai/docs/proxy/models). An OpenAI-compatible request format does not mean every OpenAI feature or parameter is supported.

#### Embeddings

```bash
curl http://127.0.0.1:8080/v1/embeddings \
  -H "Authorization: Bearer $MAPLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"nomic-embed-text","input":"Text to search later"}'
```

Embeddings use input tokens only. API usage and plan pricing are separate; see [rates](https://www.trymaple.ai/pricing).

#### Limits and failures

Request bodies are limited to **50 MiB**. The model's token/context limit is separate. Timeouts default to **300 seconds** for a response to start or a non-streaming response to finish, and **300 seconds** between streaming chunks. Change them in [proxy settings](https://www.trymaple.ai/docs/proxy/configuration#settings).

Check HTTP status before parsing an error: some failures have an OpenAI-style JSON `error`, while routing and oversized-body failures can have an empty or plain-text body. A stream can fail after its response starts. Use [troubleshooting](https://www.trymaple.ai/docs/proxy/troubleshooting) for keys, quotas, model errors and timeouts.

### System One decisions API

> Ask typed questions about private text and images and get calibrated probabilities from a model inside Maple's secure enclave. Jev-compatible schema.

Source: <https://www.trymaple.ai/docs/proxy/system-one> · Last updated: 2026-10-11

`POST /v1/systemone` answers typed questions about a JSON state, and up to four images, with calibrated probabilities instead of generated text: one masked token per question, read from GLM-5.3 Flash inside a secure enclave. The request and response follow the System One schema TypeSafe's Jev introduced, so clients built for it work against Maple by changing the base URL and model. Use it where you would otherwise write a classifier, a router or an `if` statement around an LLM, and where the state is data you do not want leaving an enclave: screens, inboxes, documents, health and money.

This API is experimental. It needs Maple Proxy 0.4.2 or newer and a [Maple API key](https://www.trymaple.ai/docs/proxy/authentication). Jev clients and SDKs that accept a base URL work when pointed at the proxy with `model` set to `glm-5-3-flash`. API usage is billed separately from plans; see [pricing](https://www.trymaple.ai/pricing).

#### Request

```bash
curl http://127.0.0.1:8080/v1/systemone \
  -H "Authorization: Bearer $MAPLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5-3-flash",
    "state": {"ticket": "Payment went through twice and I need one refunded today."},
    "questions": {
      "is_urgent": {"type": "noul", "instructions": "Does the customer need help today?"},
      "intent": {
        "type": "choice",
        "instructions": "What does the customer want?",
        "criteria": {"refund": "money back", "question": "information", "praise": "thanks"}
      },
      "frustration": {
        "type": "score",
        "instructions": "How frustrated is the customer?",
        "criteria": ["Low", "Medium", "High"]
      }
    }
  }'
```

`state` is any JSON; it is embedded in the prompt as given. Questions are answered in the order you list them, and choice options are numbered in the order of their keys. Add `"images": ["data:image/png;base64,..."]` to show up to four images with the state. `"temperature": 1` returns raw probabilities instead of calibrated ones.

```json
{
  "id": "so_7f3c…",
  "model": "glm-5-3-flash",
  "answers": {
    "is_urgent": {"type": "noul", "noul": 0.94, "temperature": 2.48, "option_mass": 1.0},
    "intent": {
      "type": "choice", "choice": "refund",
      "probabilities": {"refund": 0.86, "question": 0.10, "praise": 0.04},
      "confidence": 0.60, "temperature": 2.41, "option_mass": 0.99
    },
    "frustration": {
      "type": "score", "score": 1.2,
      "legend": {"0": "Low", "1": "Medium", "2": "High"},
      "probabilities": {"0": 0.12, "1": 0.56, "2": 0.32},
      "confidence": 0.31, "temperature": 2.48, "option_mass": 0.99
    }
  },
  "usage": {"input_tokens": 612, "output_tokens": 3, "cached_tokens": 0, "requests": 3}
}
```

#### Question types

| Type     | Answer                                                                                                            | Limits                               |
| -------- | ----------------------------------------------------------------------------------------------------------------- | ------------------------------------ |
| `noul`   | `noul`: probability that the statement is true. Optional `criteria: {"true": …, "false": …}` describes each side. |                                      |
| `choice` | `choice`: the most probable option, plus `probabilities` per option and `confidence`.                             | 2 to 255 options                     |
| `score`  | `score`: the expected level (0-based), plus `legend`, `probabilities` per level and `confidence`.                 | 2 to 10 ordered levels, lowest first |

Up to 64 questions per request; images up to 4 MiB each and 8 MiB in total. Every answer carries the calibration `temperature` applied and `option_mass`, the share of the model's next-token probability that fell on the offered options. A low `option_mass` means the question did not fit the state well; check it before trusting a confident-looking answer.

#### Errors

Schema problems return **422** and size limits **413**, each with a fixed message and an `x-opensecret-error-code` header such as `system_one_too_many_questions`, `system_one_bad_option_count` or `system_one_state_too_large`; the same code is in the body's `error.code`. Quota, plan and capacity failures use the same statuses and codes as [chat](https://www.trymaple.ai/docs/proxy/troubleshooting). A request that fails part-way returns no partial answers.

#### Cost and speed

Usage is billed as GLM-5.3 Flash tokens: one upstream request per question (two for a choice with more than 128 options), each carrying the state and the question, with one output token. A text question on a short state is about 200 input tokens. Measured on a 1080p screenshot with three questions: 1.2 to 2.9 seconds per request and about $0.003, because the image prefix is cached after the first question.

#### Privacy

The state, the images and the answers are decrypted only inside Maple's attested enclave and the provider's confidential-computing enclave, and are never stored. See [how Maple's enclaves work](https://www.trymaple.ai/proof).

### Connect OpenCode

> Connect OpenCode Desktop to Maple Proxy with a custom OpenAI-compatible provider, a Maple API key and a tool-capable model, with a link to CLI setup.

Source: <https://www.trymaple.ai/docs/proxy/guides/opencode> · Last updated: 2026-10-08

[Start Maple Proxy](https://www.trymaple.ai/docs/proxy). For coding agents, choose a [model with tool support](https://www.trymaple.ai/docs/proxy/models). The Maple desktop app also has a filled-in guide under **Local Proxy → Connect your tools**.

#### OpenCode Desktop

1. [Create a Maple API key](https://www.trymaple.ai/docs/proxy/authentication#create-an-api-key).
2. In OpenCode, open **Settings → Providers → Custom Provider → Connect**.
3. Enter these values, then connect and select the model:

| Field        | Value                                               |
| ------------ | --------------------------------------------------- |
| Provider ID  | `maple-local`                                       |
| Display name | `Maple (local proxy)`                               |
| Base URL     | `http://127.0.0.1:8080/v1`                          |
| API key      | The Maple key you just created                      |
| Model ID     | `gpt-oss-120b`, or another tool-capable Maple model |
| Model name   | Any label you like                                  |
| Headers      | Leave blank                                         |

Use a real key: a placeholder overrides the key saved by Maple. When rotating a key, disconnect and reconnect the provider.

#### OpenCode CLI

Follow OpenCode's [custom provider setup](https://opencode.ai/docs/providers/#custom-provider). Use the same base URL, real Maple key and model ID above with its OpenAI-compatible provider. For request failures, see [troubleshooting](https://www.trymaple.ai/docs/proxy/troubleshooting).

### Connect Goose

> Connect Goose Desktop or CLI to Maple Proxy, using a real Maple API key and a tool-capable model. Choose an image-capable model when sending screenshots.

Source: <https://www.trymaple.ai/docs/proxy/guides/goose> · Last updated: 2026-10-08

[Start Maple Proxy](https://www.trymaple.ai/docs/proxy), [create a key](https://www.trymaple.ai/docs/proxy/authentication#create-an-api-key), and choose a [tool-capable model](https://www.trymaple.ai/docs/proxy/models).

#### Goose Desktop

1. Open **Settings**, then the **Models** tab, and select **Configure providers**.
2. Select **Add Custom Provider**.
3. Choose the OpenAI-compatible type, and enter:
   - **API URL:** `http://127.0.0.1:8080/v1`
   - **API key:** your Maple API key
   - **Models:** the Maple model IDs you want to use
4. Select **Create Provider**, then switch to one of its models.

#### Goose CLI

The CLI's `OPENAI_HOST` is the host **without `/v1`**:

```bash
export OPENAI_HOST=http://127.0.0.1:8080
export OPENAI_API_KEY=your-maple-api-key
export GOOSE_PROVIDER=openai
export GOOSE_MODEL=gpt-oss-120b
goose session
```

For screenshots, choose a model with image support. The proxy's whole request-body limit is 50 MiB. If Goose's settings change, see its [provider guide](https://github.com/block/goose/blob/main/documentation/docs/getting-started/providers.md); for request errors, see [troubleshooting](https://www.trymaple.ai/docs/proxy/troubleshooting).

### Troubleshooting

> Resolve Maple Proxy connection, key, quota, model, browser and timeout problems using local checks, HTTP status codes and actionable error details.

Source: <https://www.trymaple.ai/docs/proxy/troubleshooting> · Last updated: 2026-10-08

#### Check the connection

Run `curl http://127.0.0.1:8080/health`. This only checks the local proxy; it does not contact Maple. If refused, start the proxy, check the host/port, publish Docker's port, or keep the desktop app open. Use `127.0.0.1` for the default address.

If desktop key creation fails, [create a key](https://www.trymaple.ai/docs/proxy/authentication#create-an-api-key) manually and set it in Local Proxy. The first inference request can take longer while the proxy verifies Maple's enclave.

#### Read the error

Inspect the HTTP status and `error.code` or `x-opensecret-error-code` when present. Some HTTP rejections have no JSON body.

| Status / detail                                                  | What to do                                                                                                                         |
| ---------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| `401`                                                            | Supply a real, unrevoked key. Remove placeholder keys. CORS mode requires a key with every request.                                |
| `400`, `unsupported_value`                                       | Check the exact model ID and supported [thinking level](https://www.trymaple.ai/docs/proxy/models#thinking-levels).                |
| `403`, `usage_limit_reached` or `free_tier_token_limit_exceeded` | Review remaining credits and plan limits in **API & credits**.                                                                     |
| `403`, `model_not_available_on_plan`                             | Choose an included model or change your plan.                                                                                      |
| Desktop `403`, browser origin                                    | Enable CORS and send a real key; see [browser access](https://www.trymaple.ai/docs/proxy/authentication#browser-clients-and-cors). |
| `404`                                                            | Check the base URL and [supported API routes](https://www.trymaple.ai/docs/proxy/api). Responses API requests are unsupported.     |
| `413`                                                            | Reduce context for `message_exceeds_context_limit`; otherwise check the 50 MiB request-body limit.                                 |
| `429` / `503`                                                    | Check [status](https://www.trymaple.ai/status), reduce concurrency and respect `Retry-After` when present.                         |
| `502`                                                            | Check network access and automatic date/time settings; update an old proxy.                                                        |
| `504` or a stopped stream                                        | Reduce the prompt or increase [request/idle timeouts](https://www.trymaple.ai/docs/proxy/configuration#settings).                  |

Retries do not guarantee a request runs only once. For persistent failures, record the proxy version and error details. Enable `MAPLE_DEBUG=true` for standalone logs if needed; remove secrets before sharing them in a [support issue](https://github.com/MaplePrivacyLabs/Maple/issues).


## Teams page summary
- Title: Maple Teams - Sovereign AI for Teams
- Value prop: Maple Teams gives businesses sovereign AI for research, writing, coding, and analysis while keeping company IP, customer data, and strategy under company control.
- Primary action: Start a Team at https://trymaple.ai/signup?next=%2Fpricing.
- Positioning: sovereign AI for business teams that need to keep IP, customer data, strategy, and internal reasoning under their control while using AI for research, writing, coding, analysis, and secure workspaces.
- Does Maple support the Model Context Protocol (MCP)? Yes. Maple supports MCP, so businesses and individuals can connect thousands of compatible data sources. Data is encrypted when it enters Maple; Maple’s encryption does not extend back into the third-party source.

## Teams sections
- Hero with Maple Teams positioning and team-plan stats
- Business ownership-risk narrative
- Interactive encryption comparison demo
- Team features
- Model Context Protocol (MCP) data-source connections
- Maple Proxy API workflow section
- Open-weight model section
- Business use-case cards
- Ownership comparison table
- Sovereign AI FAQ
- Team plan pricing
- Final CTA

## Teams feature highlights
- End-to-end encrypted workspaces: Keep strategy, customer data, research, and files inside your organization’s AI workspace.
- TEE-hosted open-weight models: Run capable open intelligence inside hardware-isolated environments that make the data boundary verifiable instead of trust-only.
- Chat, research, coding, and file creation: Give your team everyday AI workflows without routing company work through closed labs.
- Connect the data your team already uses: Maple supports MCP, giving your team access to thousands of compatible data sources. Incoming data is encrypted when it enters Maple.
- Encrypted API access: Maple Proxy API brings the same ownership boundary into workflows and developer tools that should not send company data to plaintext providers.
- Maple Proxy API: Maple Proxy API lets teams route existing apps, developer tools, automations, and internal pipelines through Maple’s encrypted infrastructure instead of sending company prompts, files, and workflow data directly to standard AI providers.
- Open intelligence is ready: Open-weight models now handle the work most teams need every day: writing, analysis, research, coding, customer insight, and decision support. Maple Teams gives your organization access to capable models without making your proprietary work part of a shared black box. Models named on the page include Kimi, GLM, DeepSeek, Gemma, GPT-OSS.

## Teams audiences and use cases
- Audience: Law firms, accounting practices, medical professionals, human rights organizations, government teams, journalists, creative studios, musicians, software teams, advisors, and operators use AI differently when their work stays theirs.
- Professional services: Draft, summarize, and analyze client work without turning privileged materials into vendor-side records.
- Creative and media teams: Develop scripts, pitches, music concepts, campaign ideas, and production plans while keeping unreleased work inside the team.
- Sensitive-mission teams: Work through source material, case notes, research, field reports, and operational plans without handing sensitive work to closed AI platforms.
- Product, finance, and operations: Explore code, forecasts, customer analysis, pricing, vendor notes, and launch plans with a sovereign AI workspace.

## Maple Team plan
- Plan: Maple Teams includes the encrypted architecture behind Maple Research, expanded for organizations that need secure workspaces, admin-managed seats, pooled usage, unified billing, MCP data connections, Maple Proxy API access, and AI workflows that keep company work under company control.
- Price: $30/user/mo.
- All features from Pro
- 2x more usage per team member
- Pool chat credits across the team
- Unified billing
- Admin-managed team seats
- MCP data source connections
- Maple Proxy API access

## Teams FAQ highlights
- What is Maple Teams? Maple Teams is a sovereign AI workspace for businesses that want AI for research, writing, coding, analysis, and file work without handing IP, customer data, or strategy to closed AI platforms.
- What does sovereign AI mean for a team? It means your organization keeps control over the work your team puts into AI: prompts, files, customer data, strategy, internal reasoning, and the value created from that work.
- Does Maple send company data to OpenAI, Anthropic, Google, or Chinese AI labs? No. Maple runs open-weight models inside protected infrastructure, including models developed by Chinese AI labs. Company prompts and files are not sent to OpenAI, Anthropic, Google, or the labs that developed those open-weight models.
- Do we need to run local AI infrastructure? No. Maple gives teams a cloud AI experience without requiring local servers, GPU management, or an internal AI infrastructure project.
- What is Maple Proxy API? Maple Proxy API lets teams route existing tools, scripts, apps, and workflows through Maple’s encrypted architecture instead of sending sensitive work directly to standard AI providers. It provides zero data retention (ZDR), so API prompts and responses are not retained.
- Does Maple support the Model Context Protocol (MCP)? Yes. Maple supports MCP, so businesses and individuals can connect thousands of compatible data sources. Data is encrypted when it enters Maple; Maple’s encryption does not extend back into the third-party source.

## Security Proof page summary
- Title: Security Proof - How Maple Works
- Value prop: Cryptographic verification, not just promises. Hardware-enforced privacy you can audit yourself, with open-source code, reproducible builds, and live enclave attestation.
- Includes an interactive encryption comparison, data-flow explanation, attestation details, privacy spectrum, security-team facts, and FAQ.
- Teams pathway: Maple Teams brings this end-to-end encrypted, verifiable architecture to company research, writing, coding, and analysis. Use Maple Proxy API to integrate Maple with existing apps and workflows, and MCP to connect thousands of compatible data sources. Explore Maple Teams at https://www.trymaple.ai/teams.

## Blog
- Index: https://www.trymaple.ai/blog
- Summary: Guides to using private, encrypted AI at work and at home, plus Maple product updates and the thinking behind confidential AI.
- RSS feed: https://www.trymaple.ai/blog/rss.xml
- RSS page: https://www.trymaple.ai/blog/feed. Get new posts in a feed reader. No email address needed, and no tracking.
- Guides filter: https://www.trymaple.ai/blog?category=guides (canonical: https://www.trymaple.ai/blog).
- Author index: https://www.trymaple.ai/blog/authors. Author profiles: https://www.trymaple.ai/blog/authors/mark-suman, https://www.trymaple.ai/blog/authors/anthony-ronning, https://www.trymaple.ai/blog/authors/michael-atwood.
- Posts are written in Ghost and published first-party on the marketing host; links to the former blog.trymaple.ai host point here.
- Model library: https://www.trymaple.ai/learn/models. Available in Maple Research means the model appears in the current model picker. API only models are available through Maple Proxy. Reference models are covered for learning and are not currently available through Maple. Model guides are Ghost posts tagged model-guide plus their exact model ID and keep their /blog/ canonical URLs.
- Posts (26, newest first):
  - GLM-5.3 Flash: What It Does and How to Use It (2026-10-07; guides, model-guide, glm-5-3-flash): https://www.trymaple.ai/blog/glm-5-3-flash-guide | Learn how GLM-5.3 Flash differs from the flagship, what its FP8 recipe requires, and how local deployment compares with private access through Maple.
  - Anthropic Reports GLM-5.3 Built Working Exploits in Cyber Tests (2026-10-07; news, glm-5-3): https://www.trymaple.ai/blog/anthropic-glm-5-3-cyber-research | Anthropic says GLM-5.3 built working V8 exploits in 50 of 410 attempts and reports weak safeguards. Here are the test conditions and NIST’s separate…
  - GLM-5.3: Capabilities, Hardware, and Access Options (2026-10-07; guides, model-guide, glm-5-3): https://www.trymaple.ai/blog/glm-5-3-guide | Compare GLM-5.3 with Flash, review the cited hardware recipe, and weigh local hosting against hosted access through Maple’s private inference path.
  - The End of Rented Intelligence: Keep Your Business AI To Yourself (2026-07-08; essays): https://www.trymaple.ai/blog/the-end-of-rented-intelligence-keep-your-business-ai-to-yourself | Some of you watched Alex Karp on CNBC last week or caught the All-In crew talking about sovereign AI over the weekend. It was a good discussion. I sat…
  - Maple for Windows: Encrypted AI, Now on Every Desktop (2026-06-30; release): https://www.trymaple.ai/blog/maple-for-windows-encrypted-ai-now-on-every-desktop | We've heard you. Today, Maple is officially available for Windows. This completes our platform lineup. Maple now runs on web, iOS, Android, macOS, Linux…
  - Meet Maple, the Personal Intelligence Platform (2026-05-06; news): https://www.trymaple.ai/blog/meet-maple-the-personal-intelligence-platform | People grow increasingly uneasy with the amount of data they share with big AI apps. You know the feeling. It's that hesitation before you type something…
  - Organize Maple with Projects, Shared Context, and Pinned Chats (2026-04-09; release): https://www.trymaple.ai/blog/organize-maple-with-projects-and-pinned-chats | If you've used Maple for a while, your chat history probably looks like mine: a giant scrolling list of every conversation you've ever had with AI. The…
  - One Year of Private AI: How Maple AI Made Encryption Easy (2026-01-16; news): https://www.trymaple.ai/blog/one-year-of-private-ai-how-maple-ai-made-encryption-easy | One year and billions of tokens later, we proved it can be done. We figured out how to give people similar power to ChatGPT, Claude, Grok, and Gemini…
  - Kimi K2 in Maple AI: Open Reasoning That Competes with ChatGPT and Claude (2025-12-30; news): https://www.trymaple.ai/blog/kimi-k2-in-maple-ai-open-reasoning-that-competes-with-chatgpt-and-claude | A new form of thinking, reasoning, and coding has come to Maple. Kimi K2 Thinking is now available in Maple’s confidential AI system. Kimi K2 is one of…
  - Maple 2.0 - Now with Live Data (2025-11-12; release): https://www.trymaple.ai/blog/maple-2-0-now-with-live-data | When Maple began less than a year ago, we set out to provide the world with the alternative to ChatGPT. Maple gave you access to high performance AI…
  - Speak Freely – Encrypted Voice Arrives in Maple (2025-09-17; release): https://www.trymaple.ai/blog/speak-freely-encrypted-voice-arrives-in-maple | What if you could talk to AI as naturally as you talk to a colleague—with the same ironclad privacy Maple delivers? Starting today, you can. We're…
  - Introducing Maple Proxy – the Maple AI API that brings encrypted LLMs to your OpenAI apps (2025-09-03; release): https://www.trymaple.ai/blog/introducing-maple-proxy-the-maple-ai-api-that-brings-encrypted-llms-to-your-openai-apps | How many apps do you come across on a daily basis that are infused with AI? Calorie trackers, CRMs, study companions, relationship apps 👀. Most of them…
  - Maple AI Model Guide With Example Prompts (2025-08-18; guides): https://www.trymaple.ai/blog/maple-ai-model-guide-with-example-prompts | 🛑NOTE: This guide is outdated. Do not use info until it has been updated for the latest models. Unlock Specialized AI Without Sacrificing Privacy At…
  - Meet Max: The Most Powerful Private AI Plan Yet (2025-07-16; release): https://www.trymaple.ai/blog/meet-maple-ai-max-the-most-powerful-private-ai-plan-yet | Just a few months ago, we launched Maple AI into a world where users felt they had no other option but to hand their data over to AI providers. People…
  - Introducing Maple Teams: Secure AI for Your Organization (2025-07-10; release): https://www.trymaple.ai/blog/introducing-maple-teams-secure-ai-for-your-organization | At Maple AI, we've always focused on delivering powerful, private AI tools that professionals can trust. Today, we're excited to introduce Maple Teams—a…
  - Get Started with a Team in Maple (2025-07-10; guides): https://www.trymaple.ai/blog/manage-your-team-on-maple | Team accounts in Maple are a powerful, secure way to use AI with your organization's confidential information. Maple allows any authenticated user to…
  - Unlock New Possibilities: Confidential Document and Image Upload in Maple AI (2025-07-03; release): https://www.trymaple.ai/blog/introducing-confidential-document-and-image-upload-in-maple-ai | Maple AI continues to revolutionize how people interact with artificial intelligence—securely and privately. Today, we’re excited to unveil a powerful new…
  - DeepSeek R1 Now Available: Enhanced AI for Maple Users (2025-06-10; release): https://www.trymaple.ai/blog/deepseek-r1-now-available-in-maple-confidential-ai | We are pleased to announce the addition of DeepSeek R1 70B as a confidential AI option in Maple AI, now available across web, desktop, and iOS for our Pro…
  - Introducing Maple AI for iPhone and iPad: Your Most Personal AI Assistant On-the-Go (2025-05-14; release): https://www.trymaple.ai/blog/introducing-maple-ai-for-iphone-and-ipad-your-most-personal-ai-assistant-on-the-go | We're excited to announce that Maple AI, the world's most private cloud AI assistant, is officially available for iPhone and iPad users around the globe!…
  - Using AI Safely in Your Therapy Practice: A Comprehensive Guide (2025-05-12; guides): https://www.trymaple.ai/blog/using-ai-safely-in-your-therapy-practice | “Confidentiality allows clients to be vulnerable and honest without fear of their information being misused.” — Kristin M. Papa, LCSW…
  - Plan Your Day Like a Pro - Time Management with Personal AI (2025-05-06; guides): https://www.trymaple.ai/blog/plan-your-day-like-a-pro-time-management-with-personal-ai | By the end of this article you will have a one-shot planning technique that organizes your day for you as if you had your own executive assistant. Why…
  - How to Move Your Personal Finance Data From ChatGPT to Maple — With This Prompt (2025-04-25; guides): https://www.trymaple.ai/blog/how-to-move-your-personal-finance-data-from-chatgpt-to-maple-with-this-prompt | AI assistants make budgeting, retirement planning, and credit-score repairs easier than ever—if you’re willing to hand them sensitive numbers. Tools such…
  - Maple AI Desktop Apps Now Available for macOS and Linux (2025-04-11; release): https://www.trymaple.ai/blog/maple-ai-desktop-apps-now-available-for-macos-and-linux | We're thrilled to announce that our Maple AI chat product is now available as a native app for macOS and Linux users! This launch marks a significant…
  - Getting Started with Maple AI: A Step-by-Step Guide (2025-04-11; guides): https://www.trymaple.ai/blog/getting-started-with-maple-ai-a-step-by-step-guide | Are you ready to experience the power of confidential and private AI? We’ve got you covered. In this article, we'll walk you through the basics of Maple…
  - Introducing the Redesigned Maple AI: Secure, Private, and Building for the Future (2025-04-02; release): https://www.trymaple.ai/blog/redesigned-maple-ai-private-ai-chat | Redesigned Maple AI web app reflects our commitment to delivering the future of private AI and dedication to the AI privacy user experience.
  - Welcome to the Maple AI Blog: Insights about Private and Confidential AI (2025-03-30; news): https://www.trymaple.ai/blog/welcome-to-the-maple-ai-blog-insights-about-private-and-confidential-ai | We're excited to announce the launch of the Maple AI blog, which will provide insights, news, and tips on how to get the most out of our groundbreaking…

## Brand guidelines
- Guidelines: https://www.trymaple.ai/brand
- Brand file (JSON): https://www.trymaple.ai/brand.json
- Name: Maple
- Legal name: Maple Privacy Labs Inc.
- Description: The personal AI you can actually trust with your home life, your work, and what you'd never tell Big AI.
- Naming: Maple: One word, capital M. Not MAPLE or maple.
- Naming: Maple AI: Only as the product descriptor, as in the Terms of Use and Privacy Notice.
- Naming: Maple Privacy Labs Inc.: When the company itself is meant.
- Core · Core color: Maple 500 #ff9771
- Core · Dark logo: Dark logo #171717
- Core · Light logo: Light logo #f7f7f9
- Supporting · Website accent: Maple 600 #f67d57
- Supporting · Secondary: Pebble 500 #8a8b9a
- Supporting · Dark ground: Pebble 800 #474854
- Supporting · Tertiary: Bark 500 #9e7469
- Supporting · Tertiary: Grove 500 #9e925e
- Typefaces: Array (display, https://www.fontshare.com/fonts/array); Manrope (body)
- Font license: Array is licensed by Indian Type Foundry under the ITF Free Font License. Get it from Fontshare. Maple does not distribute the font files.
- Wordmark: https://www.trymaple.ai/brand/maple-wordmark.svg (dark svg), https://www.trymaple.ai/brand/maple-wordmark.png (dark png), https://www.trymaple.ai/brand/maple-wordmark-light.svg (light svg), https://www.trymaple.ai/brand/maple-wordmark-light.png (light png)
- Mark: https://www.trymaple.ai/brand/maple-mark.svg (dark svg), https://www.trymaple.ai/brand/maple-mark.png (dark png), https://www.trymaple.ai/brand/maple-mark-light.svg (light svg), https://www.trymaple.ai/brand/maple-mark-light.png (light png)
- Abbreviated wordmark: https://www.trymaple.ai/brand/maple-wordmark-abbreviated.svg (dark svg), https://www.trymaple.ai/brand/maple-wordmark-abbreviated.png (dark png), https://www.trymaple.ai/brand/maple-wordmark-abbreviated-light.svg (light svg), https://www.trymaple.ai/brand/maple-wordmark-abbreviated-light.png (light png)
- App icon: https://www.trymaple.ai/brand/maple-app-icon.svg (app icon svg), https://www.trymaple.ai/brand/maple-app-icon.png (app icon png), https://www.trymaple.ai/brand/maple-icon.svg (small icon svg), https://www.trymaple.ai/brand/maple-icon.png (small icon png)
- Gradient: Maple 500 #ff9771 at 0%, Midpoint #ce9a8e at 50%, Pebble 400 #9c9dab at 100% (radial, from the bottom center)
- Clear space: Leave empty space around the logo equal to the height of the M, on every side. Do not set type, a border, or another mark inside that space.
- Do not stretch it. Do not squash it or stretch it to fill a space.
- Do not recolor it. Use the dark file or the light file. Do not pick a new color.
- Do not add effects. No strokes, shadows, outlines, or filters.
- Do not lose the contrast. Dark logo on light backgrounds. Light logo on dark backgrounds.
- Logo use: Use it to refer to Maple, in the files supplied here.
- Logo use: Do not use it in a way that suggests Maple sponsors or endorses your product.
- Logo use: Do not make it part of your own name, logo, or product.

## Status page summary
- Title: Maple status
- Public status page: https://www.trymaple.ai/status
- What it shows: Current Maple service status, component uptime, 90-day history, maintenance, and previous incidents, read from Maple's independently hosted monitor.
- Layout: one page with the overall state, any announcement, each component with its uptime and daily history, maintenance, then previous incidents. There are no per-incident URLs.
- Freshness: the page reads https://status.trymaple.ai/index.json when the site is built and again in the visitor's browser.
- Independent monitor: https://status.trymaple.ai stays the independently hosted monitor and is the source of the data.
- Snapshot when this file was built (Oct 11, 2026, 01:49 UTC): All systems operational.
- Component Maple API Health Checks (Maple Backend): Operational · 100% uptime; 90 days · no downtime.
- Maintenance: No maintenance is scheduled or in progress, and none was recorded in this history window.
- Previous incidents: 6 resolved incidents listed; the most recent started Sep 13, 2026, 19:40 UTC and lasted 21 min.

## About and legal pages
- About Maple: Maple makes capable open-source AI easy to use, private by design, and independently verifiable. Meet the people and principles behind Maple Privacy Labs.
- AI you can use with confidence. Maple Privacy Labs builds private AI products for people, teams, and developers. We make capable open-source models easy to use while giving you a way to check how your data is protected. Our goal is to put more of the best open models within reach. Privacy and proof are part of that experience: your messages are encrypted on your device, processed inside hardware-isolated infrastructure, and never used to train models.
- What Maple does: Maple Research (https://www.trymaple.ai/research), Maple Agent (https://www.trymaple.ai/agent), Maple Teams (https://www.trymaple.ai/teams), Maple Proxy (https://www.trymaple.ai/proxy-api).
- What makes Maple different: Access to open models; Privacy built into the path; A claim you can check. Details: https://www.trymaple.ai/proof.
- Who uses Maple? For individuals: Ask the difficult question. Bring the personal document. Work with capable open models in a space built to keep your conversations private. For teams: Give people room to research, write, and analyze with AI while keeping company work under company control. For developers: Bring private inference into the apps and workflows you already use through an OpenAI-compatible endpoint.
- Our Team: Michael Atwood, Head of Growth. Profile: https://www.trymaple.ai/blog/authors/michael-atwood.
- Startup programs: NVIDIA Inception, Google Cloud for Startups.
- Privacy Notice: How Maple Privacy Labs Inc. collects, stores, uses, and shares information when you use our Services.
- Terms of Use: Terms of Use governing access to Maple AI, OpenSecret, and other Maple Privacy Labs Inc. products.

## Research sections
- Hero
- Showcase
- Use-case grid
- Capability cards
- Privacy comparison table
- Plans and apps links
- Final CTA

## Pricing page summary
- Title: Maple Pricing
- URL: https://www.trymaple.ai/pricing
- Summary: Privacy is built in. Pick the usage that fits.
- Free: $0/mo. Includes: End-to-end encryption; Access to basic AI features. Not included: Web Search.
- Pro: $20/mo. Includes: All AI models including Kimi, GLM, DeepSeek, and GPT-OSS; Image and Document Upload (Apps only); Voice Recording; API Access.
- Max: $100/mo. Includes: All features from Pro; 10x more usage than Pro; Priority support; Early Access to features and models.
- Team: $30/user/mo. Includes: All features from Pro; 2x more usage per team member; Pool chat credits among team; Unified billing.
- Per-model API rates: https://www.trymaple.ai/pricing#api. USD equivalent per 1 million tokens.

## Downloads page summary
- Title: Download Maple
- URL: https://www.trymaple.ai/downloads
- Mobile: App Store at https://apps.apple.com/us/app/id6743764835; Google Play at https://play.google.com/store/apps/details?id=cloud.opensecret.maple; Play beta at https://play.google.com/apps/testing/cloud.opensecret.maple; Direct APK at https://updates.trymaple.ai/download/research/stable/android
- macOS: Download at https://updates.trymaple.ai/download/research/stable/macos
- Windows: Download at https://updates.trymaple.ai/download/research/stable/windows
- Linux: AppImage at https://updates.trymaple.ai/download/research/stable/linux-appimage; .deb at https://updates.trymaple.ai/download/research/stable/linux-deb; .rpm at https://updates.trymaple.ai/download/research/stable/linux-rpm
- Web App: Chat now at https://trymaple.ai

## Model pricing
- URL: https://www.trymaple.ai/pricing
- Unit: USD equivalent per 1 million tokens.
- GLM-5.3 Flash (glm-5-3-flash; Maple Research): $0.80 input; $0.20 cached input; $2.50 output.
- GLM 5.3 (glm-5-3; Maple Research): $3.60 input; $0.90 cached input; $11.50 output.
- Kimi K3 (kimi-k3; Maple Research): $8.00 input; $1.60 cached input; $25.00 output.
- DeepSeek V4.1 Flash (deepseek-v4-1-flash; Maple Research): $1.30 input; $0.26 cached input; $2.90 output.
- OpenAI GPT-OSS 120B (gpt-oss-120b; Maple Research): $0.30 input; standard input price (no discounted cached-input rate); $1.20 output.
- Gemma 4 31B (gemma4-31b; Maple Research): $0.80 input; standard input price (no discounted cached-input rate); $2.00 output.
- Llama 3.3 70B (llama3-3-70b; Maple Research): $3.50 input; standard input price (no discounted cached-input rate); $5.50 output.
- OpenAI GPT-OSS Safeguard 120B (gpt-oss-safeguard-120b; API only): $0.30 input; standard input price (no discounted cached-input rate); $1.20 output.
- Cached-input rates apply only to tokens reported as cached. — means the standard input price applies.
- Automatic model aliases are billed at the rate of the model selected for the request.

## Privacy architecture pillars
- Open Models: Open models you can audit. Your data isn't shared with AI companies.
- Encrypted Enclaves: Trusted Execution Environments isolate your conversations during AI processing.
- Yours Alone: Your conversations remain under your control. They stay end-to-end encrypted.

## Comparison highlights
- How conversations are protected: Maple = End-to-end encrypted; Industry standards = Provider-controlled data boundary
- Used to train AI: Maple = No, by design; Industry standards = Yes, by default
- Shared with advertisers: Maple = No, by design; Industry standards = Common practice
- Long-term memory: Maple = Coming with Maple Agent; Industry standards = Resets each chat
- How the AI works: Maple = Open and inspectable; Industry standards = Hidden inside their company

## Current navigation state
- Agent: https://www.trymaple.ai/agent
  - Status: live
- Research: https://www.trymaple.ai/research
  - Status: live
- Teams: https://www.trymaple.ai/teams
  - Status: live
- Proxy: https://www.trymaple.ai/proxy-api
  - Status: live
- How It Works: https://www.trymaple.ai/proof
  - Status: live
- Pricing: https://www.trymaple.ai/pricing
  - Status: live
- Download: https://www.trymaple.ai/downloads
  - Status: live

## Landing footer information architecture
### Products
- Maple Agent: https://www.trymaple.ai/agent
- Maple Research: https://www.trymaple.ai/research
- Maple Teams: https://www.trymaple.ai/teams
- Maple Proxy: https://www.trymaple.ai/proxy-api
- Pricing: https://www.trymaple.ai/pricing

### Resources
- Docs: https://www.trymaple.ai/docs
- Blog: https://www.trymaple.ai/blog
- RSS: https://www.trymaple.ai/blog/feed
- Model library: https://www.trymaple.ai/learn/models
- Downloads: https://www.trymaple.ai/downloads
- How It Works: https://www.trymaple.ai/proof
- Status: https://www.trymaple.ai/status

### Connect
- Contact: mailto:support@trymaple.ai
- GitHub: https://github.com/MaplePrivacyLabs/Maple
- Community: https://discord.gg/ch2gjZAMGy

### Company
- About Us: https://www.trymaple.ai/about
- Careers: https://www.trymaple.ai/careers
- Brand: https://www.trymaple.ai/brand
- Privacy Policy: https://www.trymaple.ai/privacy
- Terms of Service: https://www.trymaple.ai/terms

## Reserved routes
- https://www.trymaple.ai/faq (intentionally unresolved / 404)
- https://www.trymaple.ai/community (intentionally unresolved / 404)
- https://www.trymaple.ai/open-source (intentionally unresolved / 404)
- https://www.trymaple.ai/resources (intentionally unresolved / 404)
- https://www.trymaple.ai/contact (intentionally unresolved / 404)
- https://www.trymaple.ai/security (intentionally unresolved / 404)

## App routing note
- Primary CTAs should continue routing users to https://trymaple.ai.
- The marketing host is https://www.trymaple.ai.