plane-client
MCP server for the Plane project management API, enabling issue retrieval, listing, commenting, state changes, and creation.
README
plane-client
A small, zero-dependency Python client, CLI, and MCP server for the Plane project-management API.
Plane is a great issue tracker, and it's self-hostable — but its REST API has a handful of sharp edges that bite anyone talking to it with ad-hoc curl or requests code: control characters in JSON responses, a per-token rate limit, 400s from apostrophes, a pagination field that lies about whether there's a next page, and server-side filters that are silently ignored. plane-client encodes those workarounds once so you don't have to rediscover them.
Everything here is standard library only — no requests, no third-party HTTP stack, nothing to audit but the code in this repo. Python 3.9+.
Install
From source:
git clone <this-repo>
cd plane-client
pip install .
Or, for a checkout you plan to edit:
pip install -e .
There are no runtime dependencies. (pandoc is used if present for faithful Markdown→HTML conversion of issue/comment bodies, and a minimal built-in converter is used otherwise — but it is never required.)
Configuration
The client is bound to a single Plane project and reads its configuration from the environment:
| Variable | Required | Meaning |
|---|---|---|
PLANE_BASE |
yes | Plane instance URL, e.g. https://plane.example.com |
PLANE_WORKSPACE |
yes | Workspace slug |
PLANE_PROJECT |
yes | Project UUID |
PLANE_PAT |
one of these two | Personal access token (value) |
PLANE_PAT_FILE |
one of these two | Path to a file containing the token |
PLANE_PAT wins if both are set. PLANE_PAT_FILE is the friendlier option for keeping the secret out of your shell history and process environment — point it at a 600-mode file and you're done.
export PLANE_BASE="https://plane.example.com"
export PLANE_WORKSPACE="your-workspace"
export PLANE_PROJECT="00000000-0000-0000-0000-000000000000"
export PLANE_PAT_FILE="$HOME/.config/plane/token"
Quickstart
Python
from plane_client.core import PlaneClient
plane = PlaneClient.from_env()
issue = plane.get("PROJ-42") # by identifier, bare number, or UUID
print(issue["name"], "->", plane.url_for(issue))
plane.comment("PROJ-42", "Looks good — shipping it. Don't @ me.")
plane.set_state("PROJ-42", "In Progress") # state resolved by name, no UUIDs
for i in plane.list(state="Todo"): # client-side filtered, fully paginated
print(i["sequence_id"], i["name"])
Every issue reference — get, comment, set_state, and friends — accepts any of the three flavours below, so you can pass whatever you have on hand.
CLI
The package installs a plane console command. Each verb maps to a client method, and each supports --json for machine-readable output.
plane get PROJ-42 --json
plane list --state "In Progress" --json
# Comment body is a positional argument; use "-" to read from stdin,
# or "@path" to read from a file.
plane comment PROJ-42 "Nice work"
plane comment PROJ-42 - < notes.md
plane comment PROJ-42 @release-notes.md
plane comments PROJ-42 --json
plane set-state PROJ-42 "Done"
plane create --title "Investigate flaky test" --body-file ./body.md
plane patch-comment PROJ-42 <comment-uuid> "edited in place"
MCP server
The package ships an MCP server (stdlib JSON-RPC over stdio) exposing five typed tools to MCP-capable AI clients such as Claude Code and Claude Desktop:
| Tool | Does |
|---|---|
plane_get |
Fetch a single issue by ref |
plane_list |
List issues, optionally filtered by state name |
plane_comment |
Post a comment |
plane_set_state |
Move an issue to a named state |
plane_create |
Create an issue |
Run it directly with:
python -m plane_client.mcp_server
Registering with Claude Code
Add an entry to the mcpServers block of your ~/.claude.json. The server reads the same environment variables as the client, passed through env — so no secret ever lives in the repo:
{
"mcpServers": {
"plane": {
"type": "stdio",
"command": "python3",
"args": ["-m", "plane_client.mcp_server"],
"env": {
"PLANE_BASE": "https://plane.example.com",
"PLANE_WORKSPACE": "your-workspace",
"PLANE_PROJECT": "00000000-0000-0000-0000-000000000000",
"PLANE_PAT_FILE": "/home/you/.config/plane/token"
}
}
}
}
Plane API gotchas this handles
This is the part worth reading. Each of these is a real behaviour of Plane's API that the client absorbs so your code doesn't have to.
-
Control characters in JSON responses. Plane echoes raw control characters back inside
description_html/comment_htmlfields.json.loadsrejects the response even though the write itself succeeded. The client strips the offending control characters before parsing — and critically, does not retry a failed parse, because the write already happened and retrying would double-apply it. -
Rate limiting (429), with a retry/don't-retry distinction. Plane enforces a per-token, per-minute budget and returns
429when you exceed it. Unlike the control-char case, a429means the request did not land, so the client backs off and retries it. The two failure modes look superficially similar but demand opposite handling; the client gets both right. -
Apostrophe / quote 400s. Request bodies assembled by string interpolation return
400whenever the text contains an apostrophe or a quote. The client builds every payload withjson.dumpsend to end, so arbitrary text — contractions, quoted strings, code snippets — goes over the wire safely. -
Cursor pagination that lies about the last page. List endpoints paginate via a cursor, and the
next_cursorfield is present even on the final page. Trustingnext_cursoralone is an infinite loop. The real "is there more?" signal is thenext_page_resultsboolean, which the client keys on (falling back to a barenextURL for endpoints that don't send it). -
Server-side filters that are silently ignored. Some list filters — notably filtering the issues list by
state— are accepted by the server and then ignored, returning the entire project instead of the subset you asked for. Rather than trust the server, the client passes the filter (in case Plane ever wires it up) and post-filters the results client-side, solist(state="Todo")returns exactly the issues in that state. -
Three reference flavours. An issue can be addressed by its UUID, by its bare sequence number (
42), or by the project identifier form (PROJ-42). All three resolve transparently. Because Plane also ignores a server-sidesequence_idfilter, resolving a bare number orPROJ-Nref means paginating the issue list and matching client-side; the client caches the resolution so you only pay for it once.
Example: a coordination loop on Plane
examples/open_engine/ is a worked reference showing
how to use this client to run a small autonomous multi-agent "coding loop" with
Plane as the coordination substrate — a state machine of custom states, a
single label as the poll gate for automated workers, comment receipts as an
audit trail, and the parent field for dependency-gated work. It is illustrative;
adapt the states and IDs to your own project.
Not affiliated with Plane
This is an independent, community-built client. It is not produced, endorsed, or supported by Plane (makeplane / plane.so). "Plane" is the trademark of its respective owner.
License
MIT.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.