Momentic Agent Skills
Teach coding agents to write, run, and repair Momentic browser and mobile tests.
- Source repo
- momentic-ai/skills
- Stars
- ★ 17
- Last updated
- 1d ago
- License
- NOASSERTION
- Primary language
- JavaScript
- FA score
- 58/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude CodeChatGPT · Claude.ai (Partial support)
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Claude Code or Codex teams that want to author and run repository end-to-end browser tests directly inside the chat session.
- Not a fit if
- Teams that need fully offline testing with no Momentic cloud dependency
- Developers not using Claude Code, Codex, Cursor, Gemini CLI, or Devin
- Teams without a Momentic account or the ability to configure an API key or OAuth
- Website
- momentic.ai/docs ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-20
- Source review
- 58/100 · Major gaps
What does this agent do, and when should you use it?
Momentic Agent Skills is a plugin and skill repository that wires Momentic end-to-end testing into coding agents such as Claude Code, Codex, Cursor, Gemini CLI, and Devin. It ships four skills — momentic-test, momentic-maintain, momentic-spec, and mo-qa — together with local MCP servers (momentic, momentic-mobile) and a hosted server at https://mcp.momentic.ai. Local servers read the project's momentic.config.yaml and working directory, exposing tools like momentic_get_artifacts so agents can inspect artifacts and validate tests. Mo sessions run through the hosted server, using Connectors or a qa CLI tunnel for private targets and qa assets for mobile app builds. A Momentic account, CLI login, or API key is required, and test execution happens on Momentic's cloud platform.
The repository is a synced bundle of plugin manifests, marketplace catalogs, and skill Markdown that injects four skills and several MCP servers into a coding agent. The local momentic and momentic-mobile servers target the working directory's momentic.config.yaml and expose tools such as momentic_get_artifacts to inspect project artifacts; the momentic-test skill uses them to create, run, and validate browser, Android, and iOS tests. momentic-maintain diagnoses failures and repairs quarantined tests, while momentic-spec uses Momentic specs to improve code correctness. The hosted momentic-hosted server authenticates via OAuth and provides momentic_search_docs and Mo session tools such as momentic_mo_list_sessions; the mo-qa skill drives web or mobile bug bashes and qa assets uploads or downloads mobile app builds. Without the plugin, npx momentic skills and npx momentic install-mcp install the skill and MCP pieces separately, and npx -y qa login handles authentication.
- Claude Code or Codex teams that want to author and run repository end-to-end browser tests directly inside the chat session.
- QA engineers maintaining large e2e suites who want an agent to diagnose failures and repair quarantined test cases.
- Developers writing Android or iOS tests and validating artifacts through the momentic-mobile MCP server.
- Product teams running exploratory Mo bug bashes on PRs or new features, including localhost targets via Connectors or a qa CLI tunnel.
- Cursor, Gemini CLI, or Devin users who prefer the CLI or extension path to install Momentic skills and MCP servers.
- ChatGPT or Claude.ai users who connect through the https://mcp.momentic.ai connector to search Momentic docs and start Mo sessions.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
npx -y [email protected] --versionnpx -y [email protected] --versionnpx momentic skillsnpx -y qa login
- Environment variables
MOMENTIC_API_KEY
How do you install or deploy this agent?
For Claude Code, add the marketplace and install the plugin:
/plugin marketplace add momentic-ai/skills
/plugin install momentic@momenticBefore the first session, download the local CLIs outside Claude Code so a cold npx install does not exceed the default 30-second MCP_TIMEOUT:
npx -y [email protected] --version
npx -y [email protected] --versionCodex registers the marketplace from the terminal:
codex plugin marketplace add momentic-ai/skills
codex plugin add momentic@momenticGemini CLI installs the extension from the public repository:
gemini extensions install https://github.com/momentic-ai/skillsWithout the plugin, install skills and MCP per product. Browser tests:
npx momentic skills
npx -y qa login
npx momentic install-mcpFor mobile tests, swap in momentic-mobile:
npx momentic-mobile skills
npx -y qa login
npx momentic-mobile install-mcpFor Mo only:
npx -y qa skills --target claude-codeCursor example:
npx -y qa login
npx momentic install-mcp --yes --client cursor
npx momentic skills --yes --target cursorLocal servers resolve credentials in order from the plugin's API key, an inherited MOMENTIC_API_KEY, and the saved CLI login; the hosted server uses OAuth.
How do you use this agent?
Restart your coding agent after installation and configure only the product you use. Browser test flow: initialize config, log in, then have the agent inspect artifacts and write a first test.
npx momentic init --yes --config-only
npx -y qa login
npx momentic whoamiIn the session, call momentic_get_artifacts on the local momentic server to confirm config path and working directory, then call momentic_search_docs on momentic-hosted and load momentic-test to write and validate tests. For mobile, substitute momentic-mobile.
Mo sessions: after restart, call momentic_mo_list_sessions to check the connection without starting paid work, then load mo-qa with your target and brief. For a new tunnel, install qa and log in:
npx -y qa login
qa whoamiTo replace the plugin's hosted URL with the CLI bridge, log in under the same OS user and add the stdio server:
claude mcp add --transport stdio momentic-hosted -- npx -y qa mcpIf plugin MCP servers are still connecting on Claude Code's first turn, wait for them:
MCP_CONNECTION_NONBLOCKING=0 MCP_CONNECT_TIMEOUT_MS=30000 claudeWhat are this agent's strengths and limitations?
- Two documented paths (plugin marketplace or plain CLI) with concrete commands for Claude Code, Codex, Cursor, Gemini CLI, and Devin.
- Clear skill separation: momentic-test, momentic-maintain, momentic-spec, and mo-qa cover authoring, repair, spec validation, and bug bashes.
- Local MCP servers and the hosted momentic-hosted server are decoupled, so agents can work on repository tests and also reach cloud sessions and docs via OAuth.
- The README flags real pitfalls: plugin vs install-mcp duplication, cold npx timeouts, and the first-turn MCP connection wait.
- Test execution depends on the Momentic cloud platform and account; an API key, CLI login, or OAuth is mandatory, so it cannot be fully self-hosted.
- Cursor is not yet in the Cursor Marketplace, so you must run install-mcp and skills manually and enable MCP sign-in separately.
- The plugin bundles every product's MCP server, and unused local servers still download on first start, adding startup overhead.
- The repository says manifests and skills are synced from upstream and that outside contributors should open issues rather than PRs, so direct edits here won't take effect.
- The license is marked NOASSERTION, so commercial redistribution requires checking the actual terms first.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Momentic Agent Skills This agent | 58 · Major gaps | Agent plugin / skillFreemium | ★ 17 | 1d ago | JavaScript | Codex · Claude Code |
| agent-qa | 58 · Major gaps | CLIFree + model costs | ★ 819 | 1d ago | TypeScript | Codex · Claude Code |
| MobileBuildMCP | 54 · Major gaps | CLIFree | ★ 6.6k | 29d ago | TypeScript | Codex · Claude Code |
| AgentCairn | 89 · Good | Agent plugin / skillFree | ★ 78 | 1d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The plugin starts local MCP servers via .mcp.json and connects to hosted https://mcp.momentic.ai; the API key is declared sensitive, and README explicitly says to keep credentials out of chat and briefs and to use environment/session secrets. Skills require approval before non-idempotent side effects (submit, purchase, delete) and user consent before stopping sessions, which are positives. However, local servers run node local-mcp.cjs and pull pinned versions like [email protected] with no visible lockfile, integrity check, or supply-chain note, so deducted; rollback is limited to stopping tunnels/sessions with no recovery guidance for external data or config changes, deducted; source attribution exists via author, homepage, repository, and privacyPolicyUrl, but publisher identity is unverified, so only a moderate score based on file content.
README and SKILL.md files are broadly consistent about product selection and CLI/MCP mapping, and give concrete handling for timeouts, connection waits, and duplicate-install conflicts, so self-consistency is decent. Dependency availability relies on external npx pulls of pinned versions and hosted services, cold start can exceed the 30-second timeout, and no offline or verification mechanism is provided, deducted; failure messages are guided by instructions to report blockers, unknown outcomes, and repeated read failures, but lack specific error codes or diagnostic tables, so moderate.
README covers Claude Code, Cursor, Gemini CLI, Codex, and Devin, and distinguishes browser, mobile, and Mo scenarios, so audience and scenario coverage is strong. Capability boundaries are stated (hosted lacks local file tools; Mo uses hosted tools rather than the momentic CLI), but some boundaries rely on external doc links, deducted; trigger precision depends on skill descriptions and product-selection rules, with mixed repos requiring a question, workable but not precise; environment fit mentions working directory, config paths, and OS user, but not Windows or restricted-permission environments, deducted.
Install notes are very detailed, with per-client commands and caveats, and information architecture is clear. Examples and FAQ are thin, mostly command snippets rather than complete examples; the license is a custom NOASSERTION license with restrictions and is not prominently explained in README, deducted; version 1.3.1 is consistent across manifests but there is no changelog or version history, deducted; maintenance responsibility only says upstream sync and file issues, without support channels or response commitments, deducted; naming is mostly stable but CLI package names and skill names mix momentic/qa, deducted.
Skill outputs are operation-oriented with workflows, commands, and validation steps, giving moderate output usability; marginal value is clear for existing Momentic users but depends on paid hosted services and external CLIs, and cost-benefit for non-users is unclear with no quantified token or time cost, deducted.
Claims in README are mostly traceable to specific files or external doc links, but many referenced files inside skills are not included in the sample and cannot be cross-checked; fact/inference separation is present where skills distinguish verified results from unverified steps, but the overall evidence sample is incomplete, so low scores.
- Local MCP servers pull pinned external CLIs via npx with no visible lockfile or integrity verification, creating supply-chain risk.
- The plugin bundles both local and hosted MCP servers, potentially widening the permission surface; README warns against duplicate installs but offers no least-privilege configuration.
- The license is a custom restrictive license; confirm terms before commercial use or redistribution.
- Publisher identity is unverified; data handling and privacy of the hosted service should be independently checked.
- Several references files cited by skills are not included in the sample, so their completeness and accuracy cannot be verified.
FAQ
Is Momentic Agent Skills free?
Is Momentic Agent Skills open source?
Is Momentic Agent Skills safe to use?
Is Momentic Agent Skills still maintained?
Who makes Momentic Agent Skills?
What is Momentic Agent Skills's official website?
Thanks — we use this to improve the page.