CLI tool
Tencent/teamai-cli avatar
Tencent/teamai-cli

TeamAI turns a shared Git repo into where your team's agent skills live

Make Every Team AI Native

5,853 stars514 forksTypeScriptNOASSERTION

At a glance

What is it?
TeamAI CLI distributes skills, rules, agents, hooks, and MCP servers to the AI clients your team already uses by pulling from a shared Git repository at every session, so onboarding a teammate becomes a write permission instead of a setup document. Two of its three capability layers are labelled beta, and the per-client feature gaps in its support table are uneven.
Who is it for?
Adopt TeamAI CLI if your team already runs several AI coding clients and wants one versioned place for skills, rules, and hooks, and if you are willing to treat a Git write permission as the real onboarding step. Do not adopt it if you need review guarantees on shared rules beyond the protected default branch the README recommends.
Can I use it commercially?
Check first. The repository uses a licence we do not classify automatically, so read its LICENSE file before any commercial use.
Is it still maintained?
Yes. The repository received new commits within the last day.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • Tencent/teamai-cli5,853 ★ · +807
Star history
Repository2026-09-282026-11-15
Tencent/teamai-cli5k5.9k

807 Stars in the last 48 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

Membership is a Git write permission, not an account

TeamAI keeps no user directory of its own in the documentation. The setup path is short: an admin creates a shared-experience repo on a git host (the README names GitHub, GitLab, GitCode, CNB, TGit, or a private Git service), gives teammates permission to push branches while leaving the default branch protected, then runs `teamai init https://github.com/your-org/your-repo`. A teammate who wants in is handed a repo URL, and the join command takes that URL as its argument. Read that as a security model rather than a convenience. The sync channel is the repository itself, so every asset change is a git push with an author and a commit, and offboarding means removing write access on a host you already operate. Audit, revert, and branch tooling come along for free. The consequence is that your team standards live in a repo everyone you invited can push branches to, with the README asking admins to keep the default branch protected.

Project scope or user scope decides how wide a teammate's rules reach

`teamai init` accepts a `--scope` switch, and the README shows both values. Without it, resources are installed under the workspace directory you are standing in, which is why the member instructions begin with `cd /path/to/my-workspace`. With `--scope user`, they are installed under `~/` instead. That one flag decides the blast radius of a standard. A project-scope member carries the rules only inside that checkout, so the next repository gets nothing until somebody runs init again. A user-scope member picks up the same assets in every session on that machine, including repositories opened later, and the files land in the home directory where they can sit next to personal agent configuration. Check how local edits to a synced rule interact with a newly published version before relying on either mode.

The tutorial path: paste one line, or install the binary yourself

The quick start asks for no installation at all. You paste one line into your AI tool, it loads the teamai skill from the repository, and the `/teamai` slash command then handles setup, joining, sharing, and the dashboard. Those commands are natural-language prompts, not subcommands with flags:

text
/teamai Help me set up TeamAI for my team from scratch
/teamai Help me join my team's TeamAI, repo URL is https://github.com/your-org/your-repo
/teamai Share my xxx skill with the team
/teamai Open the TeamAI dashboard

The command-line install sits inside a collapsed details block, and it is the only route to the binary:

bash
npm install -g teamai-cli

With the binary in place, an admin or a solo user initializes against the shared repo, and each member picks a scope:

bash
cd /path/to/my-workspace
teamai init https://github.com/your-org/your-repo
teamai init https://github.com/your-org/your-repo --scope user

For teams with no repo yet, the README points at a template route: fork from the teamai-hub organisation, then run `teamai init` against the fork.

Cursor and Copilot CLI have no models column, OpenCode has no dashboard

The support table is the honest part of this project. Claude Code, CodeBuddy, WorkBuddy and OpenCode carry a check in every column. Cursor and GitHub Copilot CLI lose the models and benchmarks columns. Pi Coding Agent loses agents, sessions and benchmarks. Codex, OpenCode and Pi Coding Agent carry a footnote marker on the rules column, Pi Coding Agent has another on env, and Copilot CLI has one on scorers. Read the row for the client your team actually runs instead of assuming a rule written for one agent behaves the same in another.

Two of the three layers still carry a beta label

The catalog splits into three layers. Team Execution is the settled one: skills, rules, culture, agents, hooks, MCP, workflows, env, and models, the assets an agent needs to behave like your team on day one. Team Context is marked beta and holds learnings, a codebase graph, and a teamwiki. Team Improvement is also marked beta and holds usage, sessions, scorers, and benchmarks. The split tells you where to place your bet. Everything you would use immediately to standardise agent behaviour sits in the unlabelled layer, while the parts that accumulate knowledge over time, the codebase graph and the learnings store, are the ones still moving. The manifest supports that reading: web-tree-sitter at 0.25.10 and tree-sitter-wasms at 0.1.13 are direct dependencies, so the codebase graph is parsed with a Tree-sitter grammar bundle rather than an indexer you supply.

The npm package ships a skills tree and an agents tree, not just a binary

Look at what the package publishes and the shape of the tool becomes clear. `bin` maps the `teamai` name to `dist/index.js`, and the `files` array ships the built JavaScript, `skills`, `skill-data`, and `agents` next to the README, CHANGELOG, and LICENSE:

json
"bin": {
  "teamai": "dist/index.js"
},
"files": [
  "dist/**/*.js",
  "skills",
  "skill-data"
]

So a global install drops a skills directory and an agents directory onto the machine, not one executable. The exclusion list is a small tell about the payload: `!**/__pycache__` and `!**/*.pyc` are filtered, which means shared content can carry Python scripts even though the CLI itself is TypeScript built by tsup. The repository root shows the same split as directories, with `skills/`, `skill-data/`, and `agents/` beside `src/`, plus docs, examples, and scripts. Since a team publishes its own catalog through this path, review what lands in the repo when somebody shares a skill with the team.

Version 0.22.0 in the manifest against v0.27.0-beta.1 on the tag

The repository manifest reads version 0.22.0 while the published tags have reached v0.29.0-beta.0 on 2026-11-02, preceded by v0.28.1 on 2026-10-28 and v0.28.0 on 2026-10-26. Release tags are running ahead of the manifest version in the checkout, so pinning by manifest field and pinning by tag are not the same operation. Releases are cut with standard-version driven by .versionrc, and the history lives in CHANGELOG.md at the root. The quality gates in package.json are explicit: `npm run build` wraps tsup, `npm run typecheck` runs tsc with no emit, and `npm run test:e2e` points at a separate vitest.e2e.config.ts, which keeps end-to-end runs out of the default `npm test`. Linting runs oxlint in type-aware mode and denies warnings. Licence is MIT in package.json with a LICENSE file at the root, even though the repository licence field reads NOASSERTION, so read the file before you redistribute anything.

Every session pulls, so one bad push reaches every member

After init, every AI session pulls the latest skills, rules, and harness updates the admins published, with no manual sync step. That single behaviour is both the value and the exposure. A rule pushed in the afternoon is live for every teammate's next session, on every supported client, on machines nobody on the team administers. The repository ships the pieces for discipline: .github/ holds workflow definitions, .coding-ci.yaml and scb-check.toml sit at the root, and docs/ with examples/ci/ show how the project expects CI checks to be wired. The README's human step is to leave the default branch protected while teammates push branches; settle reviewers and how to roll back a bad rule for members whose sessions already pulled it before rollout. The last push landed on 2026-11-02 and the repository is not archived, so expect the defaults to keep moving while you decide that policy on your own host.

Editorial conclusion

Adopt TeamAI CLI if your team already runs several AI coding clients and wants one versioned place for skills, rules, and hooks, and if you are willing to treat a Git write permission as the real onboarding step. Do not adopt it if you need review guarantees on shared rules beyond the protected default branch the README recommends. Before granting access, check the support row for the client your team actually runs, because models, agents, sessions or benchmarks are missing for several of them, and pin by release tag rather than by the manifest version.

Frequently asked questions

What does the TeamAI CLI actually sync between teammates?

Skills, rules, MCP servers, and other agent resources, stored in a shared-experience repo on your own git host. Once init has run, every AI session pulls the latest version admins published, so no manual sync step is needed.

Do I need npm install for TeamAI, or does the skill alone work?

The quick start pastes one line into your AI tool to install the teamai skill, which handles setup, joining, sharing, and the dashboard. The npm install -g teamai-cli path is what gives you the teamai binary that teamai init runs.

Which AI coding clients does the TeamAI CLI support?

The support table lists Claude Code, Codex, Cursor, CodeBuddy, WorkBuddy, GitHub Copilot CLI, OpenCode, Pi Coding Agent, Oh My Pi and more. Coverage is uneven: Cursor and GitHub Copilot CLI have no models or benchmarks support, and Pi Coding Agent lacks agents, sessions and benchmarks.

Are TeamAI's codebase graph and teamwiki stable yet?

No. Team Context, which holds learnings, the codebase graph, and the teamwiki, is labelled beta, as is Team Improvement with usage, sessions, scorers, and benchmarks. The unlabelled Team Execution layer covers skills, rules, culture, agents, hooks, MCP, workflows, env, and models.

How does a new teammate join an existing TeamAI team?

They run teamai init with the shared repo URL, either inside the workspace directory for the default workspace scope or with --scope user to install under the home directory. The README asks admins to give teammates permission to push branches while leaving the default branch protected.

What licence is the TeamAI CLI published under?

package.json declares MIT and a LICENSE file sits at the repository root, while the repository licence field itself reads NOASSERTION. The package publishes under the name teamai-cli with the bin name teamai pointing at dist/index.js.

Official sources

  1. Issues
  2. README
  3. Releases
  4. Tencent/teamai-cli on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/tencent-teamai-cli.svg)](https://hysenlabs.com/projects/tencent-teamai-cli)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/tencent-teamai-cli.svg)](https://hysenlabs.com/projects/tencent-teamai-cli)

Was this useful?