> ## Documentation Index
> Fetch the complete documentation index at: https://productlane.mintlify.site/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server

> Connect Productlane to any MCP client with a v2 API key.

Productlane ships a [Model Context Protocol](https://modelcontextprotocol.io) server so AI clients (Claude, Cursor, VS Code, and any MCP-compatible app) can read and act on your workspace directly. Every v2 endpoint is exposed as a tool, so the agent can list threads, send messages, file issues, publish changelogs, and manage docs without you writing any glue code.

## Endpoint

```
https://productlane.com/api/mcp
```

The server speaks [Streamable HTTP](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports). One endpoint handles everything: `POST` for JSON-RPC calls, `GET` for the optional notification channel.

## Authentication

The MCP server uses the same bearer token as the rest of the v2 API. There's no separate login - mint a v2 key and pass it as the `Authorization` header.

1. Open **[Settings → Integrations → API](https://productlane.com/settings/integrations/api)**.
2. Click **Create API key** and pick the scopes you want the agent to have.
3. Copy the secret (`pl_v2_...`). It's shown once.

The key's scopes decide which tools the agent can call. A read-only key (`threads:read`, `docs:read`, ...) exposes only the read tools; a key without `admin` hides member management. Mint a narrow key when you want the agent limited to a subset of your workspace. See [Authentication](/api-v2/authentication) for the full scope list.

v1 keys (`pl_v1_*`) are rejected with `410 unsupported_key_version`. Mint a v2 key.

## Connect a client

### Claude (Desktop, Code, web)

Add Productlane as a custom connector and paste your key as the bearer token.

```json theme={null}
{
  "mcpServers": {
    "productlane": {
      "url": "https://productlane.com/api/mcp",
      "headers": {
        "Authorization": "Bearer pl_v2_..."
      }
    }
  }
}
```

### Cursor

`Settings → MCP → Add new MCP server`, then use the same `url` + `Authorization` header shown above.

### Any MCP client

Point the client at `https://productlane.com/api/mcp` over Streamable HTTP and send `Authorization: Bearer pl_v2_...` on every request. The server is stateless, so each request re-authenticates against the key; there's no session to keep alive.

## What the agent can do

Tools are generated from the v2 API, so the catalogue tracks the [endpoint reference](/api-v2/introduction) one-to-one. Tool names are the resource and action in `snake_case`:

| Resource   | Example tools                                                           |
| ---------- | ----------------------------------------------------------------------- |
| Threads    | `threads_list`, `threads_get`, `threads_send_message`, `threads_update` |
| Contacts   | `contacts_list`, `contacts_create`, `contacts_update`                   |
| Companies  | `companies_list`, `companies_create`, `companies_update`                |
| Projects   | `projects_list`, `projects_create`, `projects_list_statuses`            |
| Issues     | `issues_list`, `issues_create`, `issues_list_workflow_states`           |
| Tags       | `tags_list`, `tags_create`, `tags_create_group`                         |
| Changelogs | `changelogs_list`, `changelogs_create`, `changelogs_broadcast`          |
| Docs       | `docs_list_articles`, `docs_create_article`, `docs_create_draft`        |
| Portal     | `portal_get_roadmap`, `portal_get_customer_portal`                      |
| Members    | `members_list`, `members_invite`, `members_update_role` (need `admin`)  |
| Identity   | `me_get`                                                                |

A tool call is forwarded to the matching v2 endpoint, so the same scopes, rate limits, plan gating, and error envelope apply. If the agent calls a tool the key lacks scope for, it gets back `403 scope_required` naming the missing scope.

## List size and pagination

List tools return up to **25 items** per call by default (lower than the REST default of 50, to keep responses inside an agent's context window). Each response carries a `page.cursor`; pass it back as `cursor` to fetch the next page, and raise `limit` up to **200** when you want larger pages. The defaults are stated in each list tool's description so the client knows to paginate rather than assume it has everything.

## Verify the connection

Ask the client to call `me_get`. It returns the workspace, key id, and granted scopes:

```json theme={null}
{
  "workspace_id": "wks_abc123",
  "api_key_id": "key_def456",
  "scopes": ["threads:read", "threads:write", "docs:read"],
  "linear_team_ids": ["lt_..."],
  "default_linear_team_id": "lt_..."
}
```

If the call fails with `401 unauthenticated`, the header is missing or the key was revoked. If it returns `410 unsupported_key_version`, you passed a v1 key.

## Rate limits

MCP tool calls count against the same per-key limits as the REST API: 1000 reads/min and 60 writes/min by default. See [Rate limits](/api-v2/rate-limits).

## Revoke access

Revoke the key in **Settings → Integrations → API**. Every connected client using it stops working immediately. To rotate, mint a new key, update the client config, then revoke the old one.