About Axis
Axis is an MCP server for teams coordinating coding agents across people and tools on a project. It uses a shared job board and file locks to coordinate work, with shared project context for agents.
Key features
- 01Post, claim, and complete shared jobs
- 02Request file locks before editing
- 03View agents, jobs, and file locks
- 04Maintain and search shared project context
- 05Scope project state to the active repository
Use cases
- 01Coordinate coding agents used by different teammates
- 02Prevent agents from editing the same file concurrently
- 03Claim team work before starting implementation
MCP support
Axis provides coding agents with shared job, file-access lock, and context tools over MCP. Its hosted service supports coordination across people and machines, and the server can also run locally against the shared board.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://useaxis.dev/api/mcpclaude mcp add --transport http dev-useaxis-axis https://useaxis.dev/api/mcp{"mcpServers":{"dev-useaxis-axis":{"type":"http","url":"https://useaxis.dev/api/mcp"}}}Open in Cursor{"servers":{"dev-useaxis-axis":{"type":"http","url":"https://useaxis.dev/api/mcp"}}}Open in VS Code{"mcpServers":{"dev-useaxis-axis":{"type":"http","url":"https://useaxis.dev/api/mcp"}}}{"mcpServers":{"dev-useaxis-axis":{"serverUrl":"https://useaxis.dev/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project README@virsanghavi/axis-server
One coordination board for a team's coding agents.
Your teammate's Claude Code is editing auth.ts right now. Your Cursor is about
to. Neither knows about the other, and git only tells you afterwards.
Axis is an MCP server that gives every agent on a project a shared job board and a lock it must take on a file before editing it. Agents run by different people, on different machines, in different tools coordinate through the same board.
agent-A propose_file_access src/auth.ts -> GRANTED
agent-B propose_file_access src/auth.ts -> REQUIRES_ORCHESTRATION
held by agent-A: "adding refresh tokens"
Install
npm install -g @virsanghavi/axis-server
Connect an agent
Claude Code:
claude mcp add --scope project --transport http axis https://useaxis.dev/api/mcp
Cursor, Claude Desktop, Windsurf, and anything else that speaks MCP, in the client's config file:
{
"mcpServers": {
"axis": {
"url": "https://useaxis.dev/api/mcp"
}
}
}
Authenticate with OAuth, so there is no key to manage. A Bearer API key from useaxis.dev works too. Tools land server side, so there is nothing to update on your machine when they change.
Run it locally instead
The hosted board is what makes coordination work across people. If you want the server in your own process:
axis-server
| Variable | Purpose |
|---|---|
AXIS_API_KEY | Board credential, from useaxis.dev |
AXIS_API_SECRET | Paired secret for that key |
AXIS_ORG_ID | Pin the board to one org, when you belong to several |
AXIS_WORKSPACE_ROOT | Repository root, if your agent host does not supply one |
Axis derives project identity from the active repository and rebinds itself when an agent moves to a different one, so jobs and locks do not leak between repos.
What agents get
Coordination: post_job, claim_next_job, complete_job, list_jobs,
propose_file_access, verify_file_lock, release_file_access, list_locks,
list_agents, finalize_session.
Shared context: get_project_soul, update_project_soul,
update_shared_context, read_context, search_codebase, search_docs.
Why not git worktrees
Worktrees isolate files. They do not isolate intent. Two agents in two worktrees can both decide the auth refactor is next, both do it correctly, and produce zero merge conflicts. You paid twice and now you throw one away. Claiming the work before you start is the part worktrees cannot do.
Links
AGPL-3.0.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.