Umtri logo

Umtri

Maps project structure, dependencies, bugs, plans, commits, and wiki content for AI coding agents.

Install Umtri

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "umtri": {
      "command": "npx",
      "args": [
        "-y",
        "umtri-mcp"
      ],
      "env": {
        "UMTRI_API_TOKEN": "YOUR_VALUE_HERE",
        "UMTRI_API_BASE": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

Automated check passed— started and listed 30 tools correctly (Today).

Overview

The Umtri MCP server gives AI coding agents a persistent map of a project's structure, connections, dependencies, and bugs through Umtri. It exposes read, write, planning, wiki, and feedback tools, plus read-only resources describing Umtri's domain rules. Use the hosted HTTP endpoint for a quick setup, or run the npm package over stdio when you need a pinned version or custom API base. Graph queries can be narrowed by subtree, layer, role, or season to avoid retrieving the entire project map.

Use cases

  • Trace the likely impact of changing a project component
  • Retrieve a scoped project graph for an AI coding session
  • Record bugs, dependencies, APIs, and structural relationships
  • Keep plans and commits aligned with the project map
  • Organize project documentation in the Umtri wiki

Key features

  • Scoped project graph queries
  • Project, node, edge, API, and bug mutations
  • Impact analysis from recorded connections
  • Plan and commit tracking
  • Hierarchical wiki management
  • Read-only domain-rule resources

Tools30

What Umtri lets your agent do. Expand a tool for its parameters.

Verified live~13.3k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 2d ago

Recent health check succeeded.

6/6 checks healthy over the last 49d

What it can do

From the 30 tools this server returned when we ran it.

Can delete or modify data · 9Handles credentials · 6Makes network requests · 2Read-only tools · 21
Network at startup3 connections
Files written at startupNone
Install scriptsNone
npm provenanceSigned build
Package age95 days
Maintainers1

Observed 2026-11-02 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 5h ago via OSV.dev · umtri-mcp (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

npm downloads1.5kPackage downloads in the last 30 days.
Last commit1mo agoMost recent push to the default branch.
Availability100%Our rolling endpoint + install checks that succeeded.
Install checkPassedOur sandbox started it and listed its tools.
Tools exposed30Callable tools this server registers over MCP.

How Umtri works

What the Umtri MCP server does

The Umtri MCP server connects an AI coding agent to Umtri, a system that records a project's structure, the relationships between its parts, and its bugs. Instead of requiring an agent to reconstruct the codebase from source during every session, it can retrieve information already recorded in the project map.

The stored model includes a tree of project components and connections between them. The server also supports impact questions: an agent can ask what may be affected by changing a component and receive a result based on recorded relationships rather than an inferred text search. Project, bug, season, event, and wiki information are available through separate tools.

How it works

The server exposes Umtri through MCP over either a hosted HTTP endpoint or a local stdio process. The hosted option uses https://mcp.umtri.io and an authorization bearer token. The local option runs the umtri-mcp npm package and can target the default Umtri API or another API base.

Graph retrieval is scoped rather than returned as one unrestricted dump. An agent can select a subtree with rootId, restrict results by layer with maxType, filter by role or season, and control how much description data is included. Ten read-only MCP resources provide rules and templates covering vocabulary, system structure, seasons, plans, commits, transplantation, and wiki organization.

Setup and configuration

For the hosted service, create an API token in the Umtri application under Settings and API Tokens. Configure an MCP client with the HTTP URL and an Authorization: Bearer ... header. The repository shows a Claude setup command and a generic MCP configuration object using the same endpoint and header format.

For local stdio execution, run the package with npx -y umtri-mcp and set UMTRI_API_TOKEN. UMTRI_API_BASE is optional and defaults to https://api.umtri.io; set it when using a different or self-hosted API endpoint. The token is not checked during setup. A connection can therefore appear successful while calls fail because the token is invalid or lacks the required scope. Read-only tokens cannot invoke write tools.

Tools and capabilities

The Umtri MCP server groups its tools by the kind of project work they support:

  • Read project data with get_graph, get_bug, get_impact, list_projects, list_bugs, list_seasons, and list_events.
  • Change the recorded model with tools for projects, nodes, edges, APIs, bugs, and seasons. Node deletion removes a subtree only when cascade: true is supplied.
  • Manage plan execution with commit_plan and record_commit.
  • Browse and maintain the wiki with list_wiki, get_wiki, list_wiki_revisions, write_wiki, and delete_wiki. Wiki pages have a parent and position, a kind, and a place in the documentation structure; writes return warnings when a page does not fit that structure.
  • Send a suggestion or defect about Umtri with send_feedback.

The write surface includes project and node creation, node updates and deletion, relationship management, API records, bug records, and season creation. Season creation takes effect only after user confirmation.

Limitations and notes

The server depends on data that has been recorded in Umtri. Its impact results follow stored connections, so incomplete project maps may limit what can be traced. Graph results are intentionally slices controlled by query options, not complete exports.

API tokens require appropriate permissions. A token with only read access will not work for mutations, and authentication problems may not be visible until the first tool call. The available setup material documents Claude usage and a generic MCP configuration format; client-specific behavior beyond those formats is not specified.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Umtri

Umtri is an MCP server that connects AI coding agents to the Umtri project map. Its main tools read and update project structure, dependencies, APIs, bugs, seasons, plans, commits, and wiki pages. It also provides scoped graph queries, impact analysis, read-only rule resources, and feedback submission.

Use the hosted HTTP endpoint without installing a package, or run `npx -y umtri-mcp` over stdio. The hosted setup needs an API token and bearer header; the local setup requires `UMTRI_API_TOKEN` and optionally `UMTRI_API_BASE`.

Set the token in the `Authorization: Bearer ...` header for the hosted endpoint, or in `UMTRI_API_TOKEN` for stdio. Tokens are generated in the Umtri app under Settings and API Tokens. Read-only tokens cannot call write tools.

Yes. Set `UMTRI_API_BASE` to the desired API endpoint when running the stdio package. If it is omitted, the server uses `https://api.umtri.io`.

Yes. The repository documents a Claude MCP command for both the hosted HTTP endpoint and local stdio execution. It also provides a generic MCP configuration object for clients that accept one.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewUmtri AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/umtri?style=directory)](https://allmcps.com/mcp/umtri)
HTML Embed
<a href="https://allmcps.com/mcp/umtri"><img src="https://allmcps.com/api/badge/umtri?style=directory" alt="Umtri on AllMCPs" /></a>