A storyboard for one developer and their AI team.
You plan the story; your AI agents — Claude Code, Devin, Gemini, Cursor, on one machine or several — pick up the scenes, work them, and hand them back for your review. TeamEngage is the board in between: everyone sees the same plan, nobody takes the same task twice, nothing is "done" until you say so.
you: plan · approve · answer · accept agents: next → claim → log / ask → submit
╲ ╱
╲──────────▶ TeamEngage board ◀───────────────╱
draft → ready → in progress → in review → done
| Board — every task, by status | Graph — dependencies as a diagram | Sync — git between machines |
|---|---|---|
![]() |
![]() |
![]() |
te approve TK-12 TK-13 # (or on the board) these tasks may be worked
te run # your coder agent implements, your reviewer checks;
# stops with "needs you: TK-13 asks …, TK-12 review ok"
te answer TK-13 "use port 5760"
te run # picks TK-13 up where it stopped
te accept TK-12 # merges its branch (under solo: auto 30m after an ok review)
te sync --leave # leaving work: hand tasks over, push plans + code
te sync # at home: pull it all, then te run carries onte wait --notify blocks until something needs you, so you can leave
te run --watch running and get pinged.
- Agents get fed, you get asked.
te runstarts the agent CLIs you configured on each ready task, sends every submission to a reviewer model, and sendsissuesback for a fix. It only stops for you: questions, accepts, new plans. - Your task files stay yours. It tracks an existing folder of Markdown
task specs in place (one
te:line per file). - No two agents on the same work. An agent claims a task before it touches code. Claims lock the task's target paths and show up on every machine.
- Nothing is done until you say so. Agents ask instead of guessing and submit with evidence; you accept or reject.
- Two machines, one workbench.
te synccarries the plans and the code (accepted merges, open task branches) between home and work, using your own git remotes. It only fast-forwards and never merges behind your back.
Let your agent install it. Paste this into Claude Code, Devin or any coding agent:
Install TeamEngage for me on this machine, following the solo fast path in https://github.com/HefnySco/teamengage/blob/master/AGENTS.md
Or by hand:
git clone https://github.com/HefnySco/teamengage.git ~/teamengage
cd ~/teamengage && npm run deploy # deps + build + `te` on PATH
te init --overlay --root ~/code/Tasks --prefix TK --repos '../*' # track a task folder; sync the repos beside it
te import ~/code/Tasks --apply # tag the task files (dry run without --apply)
$EDITOR ~/.teamengage/machine.yaml # uncomment run: — your coder + reviewer CLIs
te run --dry-run # what it would start
te ui # the boardAny te command starts the daemon. npm run autorun keeps it running
across logins (a systemd user service).
Second machine: make the task folder a git repo with a private remote
and push it. On the other machine, clone it and the code repos, run
te init --overlay --root <same folder>, then te sync. See
team/README.md for the agents.
Setting it up on a machine, or asking an agent to? Follow AGENTS.md — step-by-step install, first/second machine, daemon service, agent rules, skills, troubleshooting. Written for AI agents, readable by humans.
More: docs/DESIGN.md (concepts and state machine) · docs/AGENTS-SETUP.md (agent protocol) · docs/TASK-FORMAT.md (task file format).
- Node.js >= 22
- git; rsync + ssh only for ssh/folder resources
npm ci # install
npm run build # compile TypeScript to dist/ and bundle the web UI
npm test # vitest
npm run lint # eslint
npm run dev # tsc --watchsrc/core pure model: schema, parser, addressing, index, state machine,
claims, validation, mermaid, event log
src/daemon single-writer daemon: store, watcher, HTTP server, sessions,
human API, SSE
src/mcp MCP Streamable HTTP server, tools, stdio shim
src/cli `te` commands
src/resources git / ssh / folder / url resource drivers
src/daemon/sync te sync: plans repo + code repos between machines, claim reconciliation
src/import importer for existing Markdown task folders
web/ human web UI, served by the daemon
skills/ agent skills: te-work (execute), te-review (peer review), te-plan (plan),
te-tag (tag domains), te-analysis (triage), te-discuss (save chats)
team/ te run setup: coder + reviewer agent CLIs
Task files in overlay workspaces follow docs/TASK-FORMAT.md (te template).
MIT — free to use, modify and redistribute, including commercially; just keep the copyright notice. See LICENSE.


