Runa MCP
OfficialThis server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
@omnidotdev/runa-mcp
A Model Context Protocol server that lets an AI agent work with Runa boards. It wraps the Runa API and authenticates as the customer (through a browser sign-in or a Runa workspace API token), exposing focused, well-described tools to list projects, columns, and tasks, read a task with its comments, and create, update, move, and comment on tasks.
Customer-facing setup and the tool reference live in the Runa docs.
Tools
Every tool except list_workspaces also takes an optional workspace (a slug
or id); see Workspaces. Tools take human identifiers:
project: the project's slug, name, task prefix (for exampleAPP), or idtask: the task's key (for exampleAPP-12) or idcolumn: the column's title (for exampleIn Progress) or id
Tool | Access | Description |
| read | The workspaces you can act in (id, slug, name) |
| read | The workspace's projects with name, slug, task prefix, and id |
| read | A project's columns (statuses) in board order, marking the done column |
| read | A project's tasks in board order, optionally by |
| read | One task with its description and comments |
| write | Create a task at the end of a column (the project's first column by default) |
| write | Change a task's title, description, priority, or due date ( |
| write | Move a task to another column of its project, at the end of that column |
| write | Comment on a task |
Plain text sent to the write tools is stored as escaped paragraphs, and stored rich text is returned as plain text.
Related MCP server: Kanban Zone MCP Server
Hosted (recommended)
Connect by URL; there is nothing to install. The hosted endpoint is:
https://mcp.runa.omni.dev/mcpSign in through the browser (default)
Add the URL to your MCP client with no credentials. On first use the client
gets a 401, discovers the sign-in server from the endpoint's protected
resource metadata, and opens your browser to sign in with your Omni account and
approve access. The agent then acts with exactly your permissions, and the
client refreshes the sign-in on its own.
{
"mcpServers": {
"runa": {
"type": "http",
"url": "https://mcp.runa.omni.dev/mcp"
}
}
}This follows the standard
MCP authorization
flow: OAuth 2.1 with PKCE, client registration through RFC 7591 dynamic client
registration, and access tokens bound to this server (RFC 8707). The sign-in
server is Gatekeeper (https://identity.omni.dev). Registered clients are
always shown a consent page and can only use https or loopback redirect URIs.
API token (CI and headless agents)
Where no browser is available, send a Runa workspace API token (runa_...,
created by a workspace admin in the workspace's settings under API tokens). It
is sent on every request and forwarded as-is to the Runa API, which applies its
scopes:
{
"mcpServers": {
"runa": {
"type": "http",
"url": "https://mcp.runa.omni.dev/mcp",
"headers": {
"Authorization": "Bearer runa_your_token_here"
}
}
}
}Tool | Scope |
|
|
|
|
|
|
|
|
Workspaces
A browser sign-in can belong to several workspaces. Call list_workspaces to
see them, then pass workspace (a slug or id) to the other tools. With exactly
one workspace it is selected automatically; with several, a tool called without
workspace returns an error asking for one. The choice is checked against your
sign-in's memberships, and the tools scope their queries to it. An API token
is bound to the workspace it was created in, so it ignores workspace.
Protocol details
The endpoint speaks MCP Streamable HTTP in stateless mode: each POST /mcp is
served independently, and GET and DELETE return 405.
GET /.well-known/oauth-protected-resource(also served at/.well-known/oauth-protected-resource/mcp) returns the RFC 9728 metadata: resourcehttps://mcp.runa.omni.dev, authorization serverhttps://identity.omni.dev, and scopesopenid profile email organization.A request without a bearer token returns
401withWWW-Authenticate: Bearer resource_metadata="https://mcp.runa.omni.dev/.well-known/oauth-protected-resource", scope="openid profile email organization".A JWT must be a Gatekeeper access token (EdDSA, issuer
https://identity.omni.dev) whose audience includeshttps://mcp.runa.omni.dev; otherwise the response is401witherror="invalid_token". Any other bearer is treated as a Runa API token and forwarded for the Runa API to validate.Tokens are never logged.
Self-hosting the HTTP server
bun run start:http # listens on 0.0.0.0:3000
docker build -t runa-mcp .
docker run -p 3000:3000 runa-mcpVariable | Default | Description |
|
| Listen port |
|
| Listen address |
|
| Runa API base URL |
|
| This server's resource identifier: the advertised |
|
| Authorization server; the required token issuer |
|
| Key set user tokens are verified against |
|
| Where |
|
| Namespace of the |
GET /health returns 200 ok for liveness and readiness probes.
RUNA_API_TOKEN is ignored in this mode; the token always comes from the
request.
Local (stdio)
Alternatively, run the server locally over stdio.
Configuration
Set two environment variables:
RUNA_API_TOKEN(required) - a Runa workspace API token, e.g.runa_....RUNA_API_URL(optional) - the API base URL. Defaults tohttps://api.runa.omni.dev.
Without a token the server still boots; every tool returns a clear error until the token is set (startup-warning convention: warn and degrade, never crash).
MCP client config
Run from a checkout with Bun, using the absolute path to src/index.ts:
{
"mcpServers": {
"runa": {
"command": "bun",
"args": ["run", "/path/to/runa-mcp/src/index.ts"],
"env": {
"RUNA_API_URL": "https://api.runa.omni.dev",
"RUNA_API_TOKEN": "runa_your_token_here"
}
}
}
}Development
bun install
bun test # unit and HTTP handler tests (no live API)
bun run typecheck # tsc --noEmit
bun run check # biome lint + format check
bun run start # run the stdio server
bun run start:http # run the hosted Streamable HTTP serverNotes and limitations
Transport: stdio (
runa-mcpbinary) and stateless Streamable HTTP (src/http.ts). The HTTP mode has no sessions, so server-initiated notifications and resumable streams are not supported.Moves:
move_taskmoves a task between columns of its own project. Moving a task to another project is done in the Runa app.
License
The code in this repository is licensed under Apache 2.0, © Omni LLC. See LICENSE.md for more information.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Task & board management for AI agents + humans. Kanban, comments, digests via MCP.
Task management for people and AI agents, with scoped OAuth access to issues, projects, and docs.
Task management for people and AI agents, with scoped OAuth access to issues, projects, and docs.
Getting Things Done (GTD) board for AI agents: capture to Inbox, next actions by context, projects, waiting-for, someday/maybe. Remote Streamable HTTP server with OAuth 2.1 login (passwordless email code). 16 tools + 4 prompts. Setup guide: https://gtdbrain.com/connect?source=glama
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables AI assistants to interact with Planka, a real-time Kanban board application, for managing projects, boards, lists, cards, and more.1044 npm6MIT

Kanban Zone MCP Serverofficial
AlicenseAqualityCmaintenanceEnables AI assistants to manage Kanban Zone workspaces via 23 tools for boards, cards, comments, checklists, and tasks.2747 npm1MIT- AlicenseBqualityDmaintenanceEnables AI agents to manage kanban boards with tasks, including creating boards, items, moving between columns, and searching.81MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to manage Kanban tasks, boards, teams, and checklists via natural language, with full CRUD operations and live updates.MIT