# Charter

> Contextual OAuth setup and launch guidance for developers and coding agents.

## Access and trust
This is a public guide and demo surface hosted on Vercel. Use the direct Markdown, REST, or stateless MCP URLs below; no Charter-specific plugin is required. Vercel deployment protection may require access before external clients can reach these endpoints.
All representations use the same Guide schema. Reviewed guides adapt to audience, capability, and stage. Saved research guides are immutable snapshots. Respect unknowns and conditional gates; missing information never means approval is unnecessary.

## Agent interfaces
- [Markdown library](https://charter-oauth-guide.vercel.app/index.md): clean catalog and machine entry point.
- [Agent workspace](https://charter-oauth-guide.vercel.app/?view=agents): MCP and REST instructions.
- [OpenAPI](https://charter-oauth-guide.vercel.app/openapi.json): HTTP schemas and operations.
- [Status](https://charter-oauth-guide.vercel.app/api/status): Research mode and storage readiness.
- [Catalog JSON](https://charter-oauth-guide.vercel.app/api/guides): Search using q and context query parameters.
- Stateless MCP: POST https://charter-oauth-guide.vercel.app/mcp. Tools: search_guides, get_guide, research_app, get_research_status.

## Reviewed guides
- [GitHub Markdown](https://charter-oauth-guide.vercel.app/guides/github.md): Choose the right app model first..
- [GitHub JSON](https://charter-oauth-guide.vercel.app/api/guides/github)
- [Gmail Markdown](https://charter-oauth-guide.vercel.app/guides/gmail.md): Reading email changes the review path..
- [Gmail JSON](https://charter-oauth-guide.vercel.app/api/guides/gmail)
- [Slack Markdown](https://charter-oauth-guide.vercel.app/guides/slack.md): Distribution is a product decision..
- [Slack JSON](https://charter-oauth-guide.vercel.app/api/guides/slack)

## Scenario parameters
audience=internal|pilot|external; capability=read|write; stage=exploring|testing|launching.
Defaults: external, read, exploring. Keep the context and guide version in implementation handoffs. This coarse capability profile does not establish exact provider scopes.

## Research
POST /api/research with {input,context}. Research executes synchronously within a bounded request and returns {job,guide?}. Read a known job using GET /api/research/{id}. Hourly limits are 8 requests per caller and 40 across the deployment; cross-origin browser research requests are rejected.
If the deployment configures RESEARCH_ACCESS_TOKEN, send Authorization: Bearer <research token> for POST /api/research and the MCP research_app tool. Read operations do not require this research token. Keep the token in the client configuration, never in guide content or URLs.
Without a configured model, live retrieval creates only a partial evidence snapshot. Never submit account credentials, callback URLs with codes, or private documentation. Model availability and provider credit must be verified in the deployed environment.