check_pause
Check if the user has paused execution. Call this before starting each node to respect user pause requests. If wait=true, blocks until the user resumes.
LABSCO SUMMARY
The plan becomes something you can edit while it is still cheap to change.
Because nodes carry fields, files and per-step instructions, the review is not just yes or no — you can pick a branch, paste in an API key and add a note before a single line is written. Note that the gate is the agent's own blocking call rather than anything standing between it and your files: pause and re-run only take effect if the agent calls check_pause and check_rerun, and get_approval itself warns that a review can take 15–30 minutes, so whatever drives it has to be willing to wait.
What it is
An approval step between a coding agent and your codebase. The agent submits its plan as XML, a local UI renders it as a flowchart you can click through, and the agent blocks on get_approval until you accept it.
What you get
- Plans submitted whole with submit_plan, or amended in place — request_plan_update takes insert, delete and replace operations against a plan already on the canvas
- get_approval blocks until you accept or cancel, and hands back the field values you filled in and the branches you chose
- Per-node progress with update_node_status — pending, active, completed, failed or skipped, plus the output that node produced
- check_pause before each node and check_rerun after the plan finishes, so a pause or a re-run request reaches the running agent
- get_resume_info returns where a paused or failed plan stopped: the current node, which nodes are done, pending or failed, and the field values and branches already chosen
- Node details read and rewritten with get_node_info and update_node_detail, or many at once with update_nodes_detail — title, description, complexity, expected output, risks
- get_usage_instructions returns plan-format instructions tailored to the agent you name: claude-code, cline, cursor, sixth or gh_copilot
- Every call takes workspace_path, so several projects can be open against one UI at the same time
- Plans auto-save to ~/.overture/history.json and reload with their field values, branch selections and attachments intact
Requirements
No account and no key. It starts as npx overture-mcp, or install it once with npm install -g overture-mcp. The UI opens on http://localhost:3031 and the WebSocket on 3030; OVERTURE_HTTP_PORT, OVERTURE_WS_PORT and OVERTURE_AUTO_OPEN change those. If the WebSocket port is already taken, Overture joins the existing server as a relay client rather than failing, which is how several agent instances share one UI. MIT.
Setup effort
One command — npx overture-mcp
USE CASES
01
Catching a misread request before the code is written
The agent posts its plan, you read the graph, and approval returns the fields and branches you settled on. Nothing runs until that call comes back approved.
02
Changing a plan that is already on the canvas
Send insert, delete or replace operations instead of a whole new plan, then ask for approval again on the amended version.
03
Watching a long run without reading a log
Each step reports its status and output as it happens, so the canvas shows what is done, what is running and what failed.
04
Picking a run back up after you paused it
Ask where it stopped — current node, completed and pending nodes, failures, and the field values and branches already chosen — and carry on from there.
05
Fixing one bad step in an otherwise good plan
Rewrite that node's title, description, expected output or risks, alone or in a batch, and re-run from it rather than starting over.
14 tools
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
check_rerun
Check if the user has requested to re-run any nodes. Call this after plan_completed to allow users to re-run specific nodes or re-run from a node to the end. Returns immediately if there's a pending request, otherwise waits briefly.
create_new_plan
Signal that you are creating a completely new, unrelated plan. IMPORTANT: Call get_usage_instructions first if you haven't already. Call this BEFORE submitting a new plan when the user asks for something unrelated to the current plan (e.g., "let's work on something else", "forget that, build X instead"). The new plan will be added alongside existing plans (Figma-style artboards). After calling this, submit the new plan using submit_plan or stream_plan_chunk, then call get_approval.
get_approval
Wait for user approval of the plan in the Overture UI. Returns status: "approved" (with field values and selected branches), "cancelled" (user rejected), or "pending" (still waiting). If status is "pending", call this tool again to continue waiting - the user may need up to 15-30 minutes to review and customize the plan.
get_node_info
Get detailed information about a specific node in the plan. Returns the node's title, type, status, description, field values, attachments, MCP servers, and branch information.
get_resume_info
Get detailed information about a paused or failed plan to help resume execution. Returns the current node where execution stopped, list of completed/pending/failed nodes, user-configured field values, selected branches, and other metadata. Use this when resuming a plan that was paused, failed, or loaded from history.
get_usage_instructions
Get detailed usage instructions for Overture MCP. CALL THIS FIRST before using any other Overture tools. Returns comprehensive documentation on how to structure plans, use XML format, handle approvals, and execute nodes. Pass your agent type to get agent-specific instructions.
plan_completed
Mark the plan as successfully completed. Call this after all nodes have been executed.
plan_failed
Mark the plan as failed. Call this if an unrecoverable error occurs during execution.
request_plan_update
Update an existing plan with incremental operations. Pass an array of operations to insert, delete, or replace nodes. Operations are applied in order with smooth animations. After calling this, call get_approval to confirm changes with the user.
submit_plan
Submit a complete plan XML to Overture. IMPORTANT: Call get_usage_instructions first to learn the correct XML format and workflow. Use this if you have the entire plan ready at once. The plan will be parsed and displayed on the canvas.
update_node_detail
Update the details of a specific node in the plan. Use this to modify the title, description, complexity, expected output, or risks of a node. The UI will update in real-time.
update_node_status
Update the status of a node during execution. Use this to show progress as you work through the plan. The node will visually update on the canvas.
update_nodes_detail
Update details for multiple nodes at once (batch operation). Use this to efficiently modify title, description, complexity, expected output, or risks for multiple nodes in a single call. More efficient than calling update_node_detail multiple times.
BEFORE YOU INSTALL
What to know first
Tools you get
14 tools when we connected. The documentation lists 11; get_node_info, update_node_detail and update_nodes_detail ship but are undocumented. The WebSocket message types listed alongside them are not MCP tools.
Requires
npx on PATH
Where it runs
All 5 documented clients run the server on your machine.
CONTEXT COST
Light Standing cost
Rated by the size of the 14 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Connect your client
Server setup
TERMINAL
npx overture-mcpMCP Clients
Claude Code
CLAUDE CODE · TERMINAL
claude mcp add overture-mcp -- npx overture-mcpCursor, Cline
JSON
{
"mcpServers": {
"overture": {
"command": "npx",
"args": [
"overture-mcp"
]
}
}
}GitHub Copilot
GITHUB COPILOT · JSON
{
"servers": {
"overture": {
"command": "npx",
"args": [
"overture-mcp"
]
}
}
}Sixth AI
SIXTH AI · JSON
{
"mcpServers": {
"overture": {
"command": "npx",
"args": [
"overture-mcp"
],
"disabled": false
}
}
}
