frontier-mcp
This MCP server exposes a markdown issue tracker (.scratch/) as a queryable graph, enabling agents to manage efforts, tickets, maps, and specs without parsing prose.
Capabilities:
List Efforts: View all efforts in a repository, including ticket counts and header docs.
Get Board: See an effort's destination, a dependency-annotated summary of its tickets, and the frontier (takeable tickets).
Get Tickets: Fetch full ticket bodies by ID.
Create Tickets: Atomically create a batch of new tickets, with temporary keys for intra-batch dependencies and cycle detection.
Update Ticket: Modify a ticket's lifecycle (claim, resolve, drop), edges, triage, comments, and acceptance criteria.
Edit Map: Read or edit an effort's map—destination, notes, fog patches—with revision protection.
Spec: Get or replace an effort's spec (opaque document) with revision protection.
Migrate Effort: Normalize legacy tickets (IDs, edges, optional rename) with a preview mode.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@frontier-mcpwhat issues are available to work on?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
frontier-mcp
An MCP server that serves the markdown issue tracker under .scratch/ as a queryable graph — so agents
stop re-parsing prose to learn what is open, blocked, or takeable.
Stability
Pre-1.0: the API can change in a breaking way on any release. The tool surface is fixed at eight
tools, but their arguments, results, and the on-disk conventions under .scratch/ are still being
settled — Ticket id format among them. Semver is honoured within 0.x as best it can be, and the
CHANGELOG.md records every break, but a 0.x bump is not a promise of compatibility. Pin an exact
version (see below) and read the changelog before moving the pin. From v1.0.0 on, breaking changes
wait for a major.
Related MCP server: brainforge-mcp
Requirements
Node 24 or later
npx to run it; pnpm for development of this package
Install once (user scope)
Register FrontierMCP once in your user-level MCP config, pinned to a released version. Every
repository you open then gets the server automatically — no per-repo mcp.json file.
In Cursor, edit ~/.cursor/mcp.json (create the file if it does not exist); other MCP clients take the
same command and arguments in their own user-scope server list:
{
"mcpServers": {
"frontier": {
"command": "npx",
"args": ["-y", "frontier-mcp@x.y.z"]
}
}
}Replace x.y.z with a published version — see the
releases. The pin is the version you get; nothing
bumps it for you.
Restart your editor after saving.
First use in a repository
Open the repository in your editor. FrontierMCP resolves the workspace from the session working directory — walking upward to the nearest
.scratch/, or.gitin either form — so opening the project is the only setup step. A git worktree is its own workspace: it carries.gitas a file, and it is served instead of the repository it was made from.Read the tracker configuration document once: fetch MCP resource
frontier://tracker-doc, or readdocs/agents/issue-tracker.mdin a repo that vendors it.Call
list_effortsto see whether.scratch/exists yet.Call
get_boardon an Effort to see the Frontier, thenget_ticketsonly for the ids you work.
A repository with no .scratch/ directory is not an error. Create the first Effort with
create_tickets, edit_map, or spec and create: true.
Override the workspace
Optional, per call or per session:
Pass
rooton any tool call to read another directory.Set
FRONTIER_ROOTin the server environment for a non-standard layout.
The working directory is the one your client launched the server in, and it is fixed for the session.
Moving to another worktree mid-session does not retarget it — use root, FRONTIER_ROOT, or restart
the server.
Tools
Eight tools: list_efforts, get_board, get_tickets, create_tickets, update_ticket,
edit_map, spec, migrate_effort. See the tracker configuration document for when to use each.
Development
pnpm install
pnpm test
pnpm run check
pnpm run build
node src/bin.tsReleasing
Publishing is CI-only via release-it. There is no local publish script, and adding one would mean minting the long-lived npm token that the CI setup exists to avoid.
Merge the work you want to ship to
master.On GitHub: Actions → Release → Run workflow, pick
auto/patch/minor/major.autois the default and derives the bump from the conventional commits since the last release —feat:gives a minor,fix:a patch. Pick an explicit one to override it. If nothing since the last tag is releasable,autoexits green having shipped nothing, so confirm a new tag appeared.The workflow runs checks + tests, bumps
package.json, updatesCHANGELOG.md, tagsv*, creates a GitHub Release, and publishes to npm withpnpm.Bump the pinned version in your user MCP config (
frontier-mcp@x.y.z) when you want the new build — pins stay manual on purpose.
The workflow is dispatchable from any branch but refuses to run off master: release-it commits,
tags and pushes as part of the run, so a release from a feature branch would rewrite that branch.
Local dry-run (no tag, no publish). It needs a clean working tree and an upstream branch, so commit first:
pnpm run release:drynpm publishing uses Trusted Publishing — GitHub Actions OIDC, configured for this repo on
npmjs.com, so there is no NPM_TOKEN to rotate.
Before changing .release-it.json or the workflow, read the release-it bullet in
AGENTS.md Work Guidance. Several settings there look removable and are not — the
absent registry-url, npm.skipChecks, npm.publishArgs, and the pnpm 10 pin each exist for a
reason recorded in one place so it cannot drift.
License
MIT — see LICENSE.
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Agent-native MCP server over the public saagarpatel.dev corpus. Read-only, stateless.
MCP server for generating rough-draft project plans from natural-language prompts.
An MCP server that gives your AI access to the source code and docs of all public github repos
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceProvides an MCP server for scraping and retrieving data from GitHub Projects, including issues, pull requests, and organizational metadata. It enables natural language interaction with project boards and repository contents using GitHub Personal Access Tokens.1
- AlicenseAqualityDmaintenanceAn MCP server that transforms markdown notes into an AI-powered knowledge graph. It enables LLM clients to explore, analyze, and diagnose knowledge graphs through tools for node explanation, path finding, causal chain analysis, and wiki health reporting.5MIT
- FlicenseNot gradedqualityCmaintenanceMCP server that enables full-text search and link navigation over Markdown files as a knowledge graph.
- FlicenseNot gradedqualityCmaintenanceMCP server for managing a project backlog as Markdown files in Git, enabling AI agents to read, create, and update tasks programmatically.2
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/51ck/frontier-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server