Agent Integration¶
Contio publishes a set of machine-readable artifacts alongside the User API docs so that coding agents and LLM toolchains can discover and call the API without hand-coding the integration.
Artifacts¶
| Artifact | URL | Purpose |
|---|---|---|
llms.txt | /user-api/llms.txt | Human- and LLM-readable index of every User API operation. Drop this URL into an agent's context window for a quick overview of available endpoints. |
operations.json | /user-api/agent/operations.json | Structured per-operation metadata (parameters, request/response schemas, auth surface). |
mcp.json | /user-api/agent/mcp.json | Model Context Protocol tool manifest for the standard user-facing operations. |
admin-mcp.json | /user-api/agent/admin-mcp.json | MCP tool manifest scoped to workspace administration operations (admin/owner only). |
Quick start
Give your coding agent the llms.txt URL and it will have enough context to explore the API, select the right endpoints, and construct valid requests:
Pointing a coding agent at the API¶
Paste the URL into the agent's prompt or context window:
Fetch operations.json for enriched per-operation metadata:
What's in llms.txt¶
The llms.txt file follows the llms.txt standard — a convention for providing LLM-readable context about a site or API. It contains:
- Header — Links to related resources (OpenAPI spec,
operations.json,mcp.json,admin-mcp.json) - Authentication summary — Operation counts per auth surface
- Operations by category — Every operation grouped by domain (Meetings, Action Items, etc.)
llms.txt is for discovery, not execution
llms.txt gives agents enough context to understand what the API can do and select the right operations. For how to call them (schemas, parameter types, auth details), agents should fetch operations.json or the OpenAPI spec.
Two MCP manifests¶
The User API ships two MCP tool manifests, split by privilege level:
mcp.json— standard user operations (meetings, action items, agenda, calendar, participants, tokens). Callable with any valid user credential.admin-mcp.json— workspace-administration operations (manage workspace settings, members, API keys, webhooks). These require a credential belonging to a workspace admin or owner.
Load the manifest that matches the privilege level your integration needs. Keep admin tools out of untrusted contexts unless the agent is acting on behalf of an administrator.
Choosing an artifact¶
| Use Case | Recommended Artifact |
|---|---|
| Quick exploration with a coding agent | llms.txt |
| Building an HTTP client or code generator | operations.json |
| Building an MCP server (user operations) | mcp.json |
| Building an MCP server (workspace admin) | admin-mcp.json |
| Full schema validation or SDK generation | OpenAPI spec |
Authentication¶
Agents still need valid credentials. The User API accepts a ContioAuth bearer token via the Authorization header. Choose a credential based on plan and use case:
- FREE workspaces: For quick, ad hoc automation, guide the user to create a short-lived access token (
cto_slt_v1_...) from Settings > API Tokens. Short-lived tokens are available on all plans, expire within hours, and are shown once at creation. See Authentication. - PRO workspaces and higher: For unattended scripts and agents, the simplest option is a Personal Access Token (
cto_pat_v1_...):
Admin MCP tools additionally require the credential to belong to a workspace admin or owner.
The auth.surface field in operations.json tells you which surface each operation requires:
| Surface | Description |
|---|---|
user-oauth | ContioAuth bearer token (OAuth token or Personal Access Token) |
admin | ContioAuth bearer token from a workspace admin/owner |
public | No authentication required (OAuth discovery, health checks) |