DeskPartner
Local shared project memory and handoff board for multiple AI assistants on Windows
- Pricing
- Open source
- Platforms
- Windows, CLI, API
- Version
- 1.0.143
- Listing updated
- today
Similar MCP servers
About DeskPartner
DeskPartner is an open-source MCP server for agent tooling. It is available on Windows, CLI, API and self-hosted.
The README describes DeskPartner as a local workspace where several AI assistants read and write the same project context, past decisions and task state. Its own pitch is that it sits beside existing clients such as Claude Desktop, Cursor and Codex rather than orchestrating models, so it neither bundles an LLM nor holds API keys.
Work is tracked through a seven-state workflow with a human gate at both ends: ideas proposed by an AI wait in a confirmation column until you accept them, and finished work goes to a review column with evidence attached. The site says the executing agent cannot mark its own work as done. Each project also carries a handoff note with revision history that a new agent can read before continuing.
Permissions are split into four risk tiers. Read-only queries pass freely, routine writes use a credential you approve once and can revoke, and changes to handoff notes, collaboration rules or deleted data require a one-time approval that shows a diff before anything is applied. Data is stored in a local SQLite database, and the README says update checks are the only network access. It targets Windows 10 and 11 only, and the README states macOS and Linux are not scheduled.
What it does
- Shared project handoff notes
- Each project keeps a Markdown handoff note with 50 revisions that new agents read before picking up work.
- Seven-state workflow with gates
- Items move through proposed, todo, in progress, awaiting review, done, suspended and discarded, with a confirmation gate at intake and a review gate at delivery.
- Tiered approval for writes
- Actions are split into four permission tiers, and high-risk changes show a before-and-after diff before you approve them.
- Local SQLite storage
- Projects, items and approval records are kept in a local database, with automatic snapshots before migrations or imports.
- Desktop quick entry
- A floating button, a global shortcut for quick notes and top-of-screen reminder banners keep the workspace reachable while you work elsewhere.
- MCP server and CLI
- Exposes its functions through a standard MCP server over stdio and a command-line interface.
Good for
- Continuing a multi-day software project after switching between AI assistants
- Splitting research or writing work so one agent drafts and another revises, with you confirming the result
- Tracking a long personal project such as exam prep or a plan, returning to it days later
- Reviewing what an AI agent claims to have finished before accepting it as done
Pricing
Open source
MIT-licensed open source; sponsorship links are optional
Prices change. Check lynnnn928.github.io before you buy.
DeskPartner FAQ
Does DeskPartner include its own AI model?
No. The README says it does not bundle a large language model, hold API keys or initiate inference, and works alongside existing AI clients.
Which operating systems are supported?
Windows 10 and 11 (64-bit). The README states macOS and Linux are not scheduled.
Where is project data stored?
By default in a local SQLite database under the user's AppData folder. The README says update checks are the only network access.
How are AI agents connected?
Through a standard MCP server configured in the client's MCP settings, or through the CLI. An agent registers itself and waits for a one-time approval in the desktop app before writing.
Is DeskPartner free?
Yes. DeskPartner is open source, so you can use and self-host it for free.
What are the best alternatives to DeskPartner?
The closest alternatives to DeskPartner are Seamless, Agent Memory MCP and HeapFile.
Where this comes from
- Description: written from the project's README, the project's website and its MCP Registry entry.