Onplana

What is Onplana?

Onplana is an AI-native project management platform, projects, tasks, sprints, milestones, risks, timesheets and portfolio governance, with a hosted MCP server so your agent can work in it directly. Connect once and Claude, ChatGPT, Cursor, Gemini or Copilot can read your portfolio, create and update work, search everything your org has written, and draft status reports, all under your own account with a full audit trail.

The server is remote (Streamable HTTP) at https://mcp.onplana.com/mcp, so there is nothing to run locally. Onplana is listed in both the ChatGPT and Claude connector directories.

How to use Onplana?

Add the remote server to any MCP client and sign in with OAuth:

{ "mcpServers": { "onplana": { "url": "https://mcp.onplana.com/mcp" } } }

• Claude and ChatGPT: add Onplana from the connector directory, or paste the URL as a custom connector. OAuth handles the rest.
• Gemini CLI: set ONPLANA_PAT, then gemini extensions install https://github.com/Onplana/onplana-mcp-server.
• Scripts and in-house agents: mint a personal access token in Settings, send it as Authorization: Bearer <token>, or use the typed onplana-mcp-client SDK.

You need an Onplana account. MCP is available on every plan, including the free one.

Key features of Onplana

• 29 tools, 16 read and 13 write: projects, tasks, sprints, milestones, comments, risks, dependencies, timesheets and team members.
• Search across the whole org: search_org_knowledge runs hybrid BM25 and vector RAG over tasks, projects, wiki pages and comments.
• AI grounded in the real plan: summarize_project, analyze_project_risks and generate_status_report synthesize from live data, not from a prompt.
• Prompt-injection containment: content your team wrote is wrapped and delimited before it reaches the model, so a planted instruction arrives as data.
• Audited and governed: every write lands under your identity with an audit trail, plan gates are enforced server-side, and destructive operations are deny-by-default per organization.
• Open source under the hood: the MIT-licensed server template and client SDK in this repo are the same transport layer Onplana runs in production.

Use cases of Onplana

• Ask what is slipping this sprint and get an answer from the live schedule instead of holding a status meeting.
• Turn a brief into a plan: create the project, tasks, milestones and dependencies in one conversation.
• Keep the plan current from wherever you work, letting the agent move dates, owners and status as things change.
• Draft the weekly status report from the actual schedule, risks and recent activity.
• Recover context: "what was the rationale for the three-week design phase?" across everything the org has written.
• Build your own MCP server from the template, with the security primitives already wired.

FAQ from Onplana

Do I need to run anything locally?

No. It is a hosted remote server over Streamable HTTP. Add the URL, sign in, done. The GitHub repo is there if you want to read the transport layer or build your own server, not because you need it to connect.

How does authentication work?

OAuth 2.1 with dynamic client registration for interactive clients, or a personal access token for scripts and CI. Tokens are scoped and bound to one organization, and any connection can be revoked from Settings.

How is prompt injection handled?

User-generated content is wrapped in delimiters before it enters the model's context, so planted text reads as data rather than instructions. That is mitigation rather than a guarantee: the real boundaries are the permission checks on every write and the audit trail behind them.

Which plans include MCP?

All of them, including Free. Individual tools can be gated to higher tiers, and the server tells the agent which plan unlocks a gated tool instead of failing silently.

Can I use the code in my own project?

Yes, it is MIT licensed. The server template needs Node 20 or newer, the client SDK Node 18 or newer, plus @modelcontextprotocol/sdk 1.29 or newer and Express 4.18 or 5.