Your agents' plan shouldn't live in one chat.
Nitka is a hosted roadmap your AI agents share. Claude, ChatGPT, Cursor, and other supported MCP clients connect to one versioned plan. Every connected agent reads and updates the same current state; you steer it from one board.
Supported remote MCP clients · No model inside · Every change versioned
Context dies with the chat.
You plan a project with an AI. The roadmap stays trapped in that conversation. Tomorrow you re-explain it to another chat, model, or device. Local files help one runtime; Nitka gives every connected agent the same durable plan.
How it works
shared context for AI agentsCreate a roadmap
From chat or from the web app: goal, stages, steps, sub-steps.
Connect your chats
Add Nitka’s remote MCP endpoint in a supported client, authenticate with OAuth, and enable it in the chats that need the roadmap.
Agents update, you steer
Connected agents update stages as they work. You steer every project from one board, and each client picks up the latest state on its next read.
One board. Every project.
product preview · example roadmappersistent plan across chats — read by every connected agent
Why nitka
Cross-vendor by design
One plan, available from supported AI clients across vendors. Not locked to one platform.
No model inside
Nitka is a deterministic state store. It does not train on or generate from your plans; connected clients receive only the roadmap data they request.
Built for agents
Compact reads return the current step and recent changes without loading the full roadmap.
Versioned everything
Every write is attributed, versioned, and diffable. You can see what changed, who changed it, and when.
Questions
What is MCP?
MCP is an open standard that lets AI apps connect to external tools and data. Nitka is a remote MCP server for project plans: connected clients can read and update the roadmap through it. More at modelcontextprotocol.io.
Which AI apps work with it?
Nitka works with supported remote MCP clients, including Claude, ChatGPT, and Cursor. Availability, write permissions, and setup vary by client, plan, and surface. Compatibility, not partnership.
Does Nitka train on my data?
No. Nitka runs no model and does not use your plans for training. It stores and versions roadmap data. When a connected AI client reads that data, the client handles it under its own privacy and retention terms.
When does it launch?
Public beta is planned for autumn 2026. Waitlist members get access first.
Give your agents a thread to follow.
Enter a valid email address.
Something went wrong — try again or email hello@nitka.sh
No spam. One email when the beta opens.