From tensorpm
Agentic Project Memory | AI Project Management Software. The dedicated project management subagent for Claude Code, Codex, OpenClaw and any MCP/A2A agent — manages shared project context (goals, action items, decisions, history) via MCP tools and the A2A protocol. Local-first, free.
How this skill is triggered — by the user, by Claude, or both
Slash command
/tensorpm:tensorpmThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill to operate TensorPM as the dedicated project management subagent — via MCP for structured CRUD and A2A for high-level, context-aware operations on a running TensorPM desktop app.
Use this skill to operate TensorPM as the dedicated project management subagent — via MCP for structured CRUD and A2A for high-level, context-aware operations on a running TensorPM desktop app.
TensorPM has two interfaces over the same project graph:
TensorPM itself is free. For AI capabilities outside MCP (A2A), use your own API key (BYOK) or create an account.
set_api_key).message/send).Preferred package-manager install:
# macOS
brew install --cask neo552/tensorpm/tensorpm
# Windows (PowerShell)
winget install --id Neo552.TensorPM --exact --accept-package-agreements --accept-source-agreements
Direct-download fallback (works on every platform, no installer script required):
# macOS DMG
curl -fL -o /tmp/TensorPM-macOS.dmg \
https://github.com/Neo552/TensorPM-Releases/releases/latest/download/TensorPM-macOS.dmg
hdiutil attach /tmp/TensorPM-macOS.dmg -nobrowse -quiet
cp -R "/Volumes/$(ls /Volumes | grep -i TensorPM)/TensorPM.app" /Applications/
hdiutil detach "/Volumes/$(ls /Volumes | grep -i TensorPM)" -quiet
# Linux AppImage
curl -fL -o ~/TensorPM.AppImage \
https://github.com/Neo552/TensorPM-Releases/releases/latest/download/TensorPM-Linux.AppImage
chmod +x ~/TensorPM.AppImage
# Windows installer
Invoke-WebRequest -Uri https://github.com/Neo552/TensorPM-Releases/releases/latest/download/TensorPM-Setup.exe -OutFile $env:TEMP\TensorPM-Setup.exe
Start-Process $env:TEMP\TensorPM-Setup.exe
If the agent needs to verify or pin a version, list releases via gh release list -R Neo552/TensorPM-Releases (the GitHub /releases/latest endpoint excludes pre-releases, so prefer gh release list when only beta tags exist).
POST /integrations/mcp/install).http://localhost:37850.When a user asks to give TensorPM access to external MCP servers, write an MCP config file and let TensorPM import it on startup.
Preferred paths:
~/.tensorpm/agent-mcps.json.tensorpm/agent-mcps.json.tensorpm/agent-mcps.local.jsonTENSORPM_AGENT_MCP_CONFIG_FILE to one or more paths separated by the OS path delimiter.TensorPM accepts standard mcpServers config blocks and TensorPM-native agentMcpServers. See Agent MCP Clients for schema, examples, and safety notes.
The UI exports its own generated snapshot to ~/.tensorpm/agent-mcps.generated.json; agents should write user-managed config to ~/.tensorpm/agent-mcps.json.
| Task | Use | Why |
|---|---|---|
| Structured action-item CRUD | MCP tools | Direct typed operations |
| Set provider API keys | MCP set_api_key | Dedicated secure write-only tool |
| Project-wide/contextual changes | A2A message/send | Managed by project manager agent |
| HTTP-based automation/client integration | A2A REST/JSON-RPC | Endpoint-first integration path |
| Multi-turn planning with conversation state | A2A with contextId | Native conversation continuity |
Rule of thumb:
list_*, get_project, or A2A read endpoint) to confirm state.YYYY-MM-DD).propose_updates requires human approval before apply.Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub neo552/tensorpm-skill --plugin tensorpm