Skip to content
Agents tracked: 492 Releases (7d): 545 Agent status: 1 with issuesSince last limit reset: Claude 9d · Codex 7d For AI agents: Markdown Updated Nov 20, 2026

Put an AI agent to work inside your GitHub

An agent can review every pull request, answer when you mention it in an issue, check AI-written code for security problems and run tests on your prompts. You do not need a server for any of it: GitHub runs the agent for you, from one small file in your repository. Below are 39 ways to do it, checked against their own repositories, and what 3,053 real projects that use AI agents actually run.

How it works

  1. You do something in GitHubopen a pull request, write @claude in an issue, or push code
  2. GitHub starts the workflowa GitHub Action, defined in a file in .github/workflows/
  3. The agent reads the codewith the API key you stored in the repository's secrets
  4. It answers or changes the codea review comment, a commit, a pull request or a failed check

The model is paid through your own API key at the vendor's prices. GitHub Actions minutes are free in public repositories; private repositories use the minutes included in your GitHub plan. (GitHub billing)

What do you want the agent to do?

Pick a job. Each card shows the most used option first, then alternatives.

What projects with AI agents actually use

Out of 7,205 public repositories set up for AI coding agents that AgentGid scanned, how many run each of these in their workflows.

Claude Code Action v1.0131
Codex Exec Action11
AI Inference6
HOL AI Plugin Scanner6
Run Gemini CLI5

Set it up in three steps

The same for every option below; the README of each Action lists its own inputs.

  1. Get an API key from the model vendor (Anthropic, OpenAI, Google) or use the GitHub token the Action asks for.
  2. Add the key in your repository: Settings, then Secrets and variables, then Actions.
  3. Add a workflow file in .github/workflows/ with the uses: line below and the inputs from the Action's README.

Mention an agent in an issue or pull request and get the change

Write @claude (or run Codex, Gemini CLI, Jules) in an issue or a pull request: the agent runs inside GitHub, reads the repository, answers, and can push a commit or open a pull request.

Keep in mind The agent reads your code and, with write permission, can push commits: give the workflow only the permissions it needs.

Option Line for your workflow Keys it asks for In agent projects GitHub stars License
Claude Code Action v1.0 Flexible GitHub automation platform with Claude. Auto-detects mode based on event type: PR reviews, @claude mentions, or custom automation. Official Action of Claude Code.
uses: anthropics/claude-code-action@v1
latest release Aug 26, 2025
anthropic_api_key, claude_code_oauth_token, github_token, ssh_signing_key 131 9,972 MIT
Codex Exec Action Run `codex exec` with a prompt. Official Action of OpenAI Codex.
uses: openai/codex-action@<version> openai-api-key 11 1,305 Apache-2.0
AI Inference Generate an AI response based on a provided prompt. The current action.yml supports only the copilot provider, which runs the GitHub Copilot CLI. The repository still describes it as calling models with GitHub Models, a service GitHub retired on July 30, 2026: check which version your workflow pins.
uses: actions/ai-inference@v3
latest release Jul 29, 2026
token 6 527 MIT
Run Gemini CLI Invoke the Gemini CLI from a GitHub Action. Official Action of Gemini CLI.
uses: google-github-actions/run-gemini-cli@v0.1.22
latest release Apr 24, 2026
gemini_api_key, google_api_key 5 2,128 Apache-2.0
Claude Code Base Action Run Claude Code in GitHub Actions workflows. Official Action of Claude Code.
uses: anthropics/claude-code-base-action@<version>
latest release May 19, 2025
anthropic_api_key, claude_code_oauth_token — 1,055 MIT
Jules Invoke Invoke Jules, an AI-powered coding agent, to perform tasks on your codebase. Official Action of Jules.
uses: google-labs-code/jules-action@v1.0.0
latest release Dec 15, 2025
jules_api_key — 268 MIT
Run an Oz cloud agent Run the Oz agent in a GitHub workflow. Official Action of Warp.
uses: warpdotdev/oz-agent-action@v1.0.37
latest release Oct 13, 2026
warp_api_key — 121 —
cagent Runner Run a cagent AI agent with a single line.
uses: docker/cagent-action@v1.5.5
latest release Jun 15, 2026
anthropic-api-key, openai-api-key, google-api-key, aws-bearer-token-bedrock, xai-api-key, nebius-api-key, mistral-api-key, github-token, org-membership-token — 19 Apache-2.0

Get an AI review on every pull request

Each new pull request gets a summary and review comments from an LLM before a person looks at it.

Keep in mind The diff is sent to the model provider you choose; the review is a suggestion, not a guarantee.

Option Line for your workflow Keys it asks for In agent projects GitHub stars License
The PR Agent Summarize, review and suggest improvements for pull requests.
uses: The-PR-Agent/pr-agent@v0.50.0
latest release Nov 7, 2026
none in its inputs — 16.3K MIT
ChatGPT CodeReviewer A Code Review Action Powered By ChatGPT.
uses: anc95/ChatGPT-CodeReview@v1.0.24
latest release Jul 14, 2026
none in its inputs — 4,476 ISC
Shippie 🚢 an extensible code review agent built on flue.
uses: mattzcarey/shippie@v0.21.2
latest release Jun 23, 2026
ANTHROPIC_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY, CLOUDFLARE_API_KEY, GITHUB_TOKEN — 2,525 MIT
OpenQodex AI Code Review: SAST, Secret Scanning and Lint for Pull Requests AI code review and scanners for pull requests: SAST, secrets, dependencies, lint. Review needs your Anthropic key.
uses: openqodex/openqodex@<version>
latest release Nov 11, 2026
none in its inputs — 1,198 Apache-2.0
AI Code Review Action Perform code reviews and comment on diffs using OpenAI API.
uses: villesau/ai-codereviewer@v2.7.0
latest release Dec 2, 2023
GITHUB_TOKEN, OPENAI_API_KEY — 1,047 MIT
GenAI Code Review Reviews pull request diffs with an LLM (Gemini, OpenAI, Anthropic or any OpenAI-compatible server) and posts only the few inline comments most worth a human's attention, saying what it held back and what it did not review.
uses: cirolini/genai-code-review@v3.0.1
latest release Sep 21, 2026
api_key, openai_api_key, github_token — 372 MIT
Presubmit - AI Code Review & Fixes AI-powered code reviews that detect issues, provide summaries, and suggest auto-fixes on PRs.
uses: presubmit/ai-reviewer@v0.2.5
latest release Nov 8, 2025
none in its inputs — 187 MIT
Juror AI Multi-model PR review that shows you the bill — duplicate findings collapse into one, with optional agreement filtering.
uses: Juror-AI/juror@v1.4.8
latest release Sep 30, 2026
github-token — 101 MIT
Sourcery Action A GitHub Action for running Sourcery. Official Action of Sourcery.
uses: sourcery-ai/action@v1
latest release Nov 9, 2022
token — 33 MIT

Check AI-written code and agent setups for security problems

Scans what agents and AI-written code bring in: leaked keys, risky agent and MCP configs, vulnerable code, before merge.

Keep in mind Scanners find patterns, not every problem; keep human review for anything that touches secrets or production.

Option Line for your workflow Keys it asks for In agent projects GitHub stars License
HOL AI Plugin Scanner Scan AI plugin repositories for security, publishability, and best-practice quality signals.
uses: hashgraph-online/ai-plugin-scanner-action@v1.2.817
latest release Nov 14, 2026
submission_token 6 10 Apache-2.0
OpenAI Codex Security Scan code with Codex Security, enforce findings policy, and produce JSON and SARIF reports. Official Action of OpenAI Codex.
uses: openai/codex-security@<version>
latest release Nov 12, 2026
none in its inputs — 13.7K Apache-2.0
Claude Code Security Reviewer AI-powered security review GitHub Action using Claude to analyze code changes for security vulnerabilities. Official Action of Claude Code.
uses: anthropics/claude-code-security-review@<version> claude-api-key — 6,924 MIT
Pipelock Agent Security Scan Scan your project for AI agent security risks. Pattern-based detection of credentials and misconfigurations with tailored config generation. See docs for coverage details.
uses: pipelab-org/pipelock@v3.8.0
latest release Nov 10, 2026
none in its inputs — 996 Apache-2.0
Skylos - Python SAST, Dead Code Detection & PR Gate SAST, dead code, secrets, PR gates, and digest-pinned container image vulnerability scans.
uses: duriantaco/skylos@v5.6.0
latest release Nov 11, 2026
token — 945 Apache-2.0
aislop — AI Code Quality Gate Run the aislop quality gate on AI-assisted code (scan + fail CI below threshold from .aislop/config.yml).
uses: scanaislop/aislop@v0.22.1
latest release Nov 14, 2026
none in its inputs — 746 MIT
Vet: Verify Everything LLM-based code review that finds issues tests and linters miss.
uses: imbue-ai/vet@v0.2.16
latest release Sep 14, 2026
none in its inputs — 493 AGPL-3.0
AI-BOM Scan Discover and inventory AI/LLM agents, models, and API integrations in your codebase. Generates SBOM-style reports in CycloneDX, SARIF, SPDX, and more.
uses: Trusera/ai-bom@v3.1.0
latest release Feb 12, 2026
none in its inputs — 363 Apache-2.0
node9 Agent Security Scan a repo for agent-security risks (injectable CI workflows, unsafe agent configs, MCP) and comment on the PR.
uses: node9-ai/node9-proxy@v2.62.1
latest release Nov 15, 2026
github-token — 225 Apache-2.0
LintLang Agent Config Linter Statically lint AI agent configs, tool definitions, and prompts with LintLang.
uses: hermes-labs-ai/lintlang@v0.8.3
latest release Oct 20, 2026
none in its inputs — 147 Apache-2.0
AI Surface Check Map AI attack surfaces in pull requests and gate new high-risk findings before merge.
uses: apisec-inc/AI-Surface@v1.1.1
latest release Oct 8, 2026
github-token — 144 MIT

Test prompts, agents and AI features on every change

Run evaluations or a mock model server in CI, so a prompt or agent change that breaks answers fails the build.

Keep in mind Evaluations that call a real model cost tokens on every run; mock servers do not.

Option Line for your workflow Keys it asks for In agent projects GitHub stars License
Setup aimock Start an aimock server for AI application testing — LLM APIs, MCP, A2A, AG-UI, vector DBs.
uses: CopilotKit/aimock@v1.48.0
latest release Nov 11, 2026
none in its inputs — 1,015 MIT
mcpsnoop Fail CI on what an MCP server actually did on the wire, and file every finding as a code scanning alert.
uses: kerlenton/mcpsnoop@v0.28.0
latest release Oct 17, 2026
none in its inputs — 368 MIT
coder_eval Install a pinned coder-eval and run evaluation tasks as a CI gate, with JUnit XML output.
uses: UiPath/coder_eval@v0.12.18
latest release Nov 9, 2026
none in its inputs — 160 Apache-2.0
Azure AI Agent Evaluation Evaluate the performance, quality and safety of AI agents using the Azure AI Evaluation SDK.
uses: microsoft/ai-agent-evals@<version> none in its inputs — 107 MIT
Test LLM outputs Run promptfoo evaluations on your prompts. Works with pull requests, pushes, and manual triggers (workflow_dispatch).
uses: promptfoo/promptfoo-action@<version>
latest release Jun 15, 2026
github-token, openai-api-key, azure-api-key, anthropic-api-key, huggingface-api-key, aws-access-key-id, aws-secret-access-key, replicate-api-key, palm-api-key, vertex-api-key, cohere-api-key, mistral-api-key, groq-api-key — 81 MIT
ollama-action Run Ollama large language models (LLMs) with GitHub Actions.
uses: ai-action/ollama-action@v2.0.1
latest release Jun 24, 2026
none in its inputs — 34 MIT
Braintrust eval Automatically run evals on your AI project using Braintrust.
uses: braintrustdata/eval-action@v2.1.0
latest release Sep 1, 2026
api_key, github_token — 21 MIT
Tessl Skill Review Automatically review SKILL.md files changed in a pull request and post results as a PR comment.
uses: tesslio/skill-review@<version> tessl-token — 19 —

Translate the app and fix accessibility automatically

Machine translation of locale files and docs, and accessibility checks that open issues and propose fixes.

Keep in mind Translations and fixes arrive as changes to review; read them before they ship.

Option Line for your workflow Keys it asks for In agent projects GitHub stars License
Lingo.Dev AI Localization Automated AI localization for dev teams.
uses: lingodotdev/lingo.dev@<version>
latest release Nov 9, 2026
api-key — 5,402 Apache-2.0
Languine Self-hosted automated localization powered by your own Languine deployment.
uses: languine-ai/languine@v1
latest release Feb 5, 2025
api-key — 2,032 MIT
accessibility-scanner Finds potential accessibility gaps, files GitHub issues to track them, and attempts to fix them with Copilot.
uses: github/accessibility-scanner@v3.5.2
latest release Nov 5, 2026
token, password — 397 MIT

For engineers: what the CI of AI-agent projects runs

The 40 Actions used most in the workflows of 7,205 public repos set up for AI coding agents (AgentGid's own scan), and how often each is pinned to a full commit SHA, as GitHub's security guide recommends. Across all uses: 26% pinned.

# Action Projects Pinned to SHA Seen with
1 actions/checkout 3008 18% claude-code, github-copilot, cursor, openai-codex
2 actions/setup-node 1901 19% claude-code, github-copilot, cursor, openai-codex
3 actions/upload-artifact 1209 26% claude-code, github-copilot, cursor, openai-codex
4 actions/setup-python 795 20% claude-code, github-copilot, cursor, gemini-cli
5 actions/download-artifact 705 31% claude-code, github-copilot, cursor, openai-codex
6 actions/cache 569 29% claude-code, github-copilot, cursor, openai-codex
7 pnpm/action-setup 471 19% claude-code, cursor, github-copilot, openai-codex
8 docker/login-action 454 26% claude-code, github-copilot, cursor, gemini-cli
9 docker/setup-buildx-action 446 25% claude-code, github-copilot, cursor, gemini-cli
10 docker/build-push-action 432 25% claude-code, github-copilot, cursor, gemini-cli
11 softprops/action-gh-release 410 18% claude-code, github-copilot, cursor, openai-codex
12 actions/github-script 378 37% claude-code, github-copilot, cursor, openai-codex
13 actions/deploy-pages 335 18% claude-code, github-copilot, cursor, openai-codex
14 actions/upload-pages-artifact 327 17% claude-code, github-copilot, cursor, openai-codex
15 oven-sh/setup-bun 299 24% claude-code, cursor, github-copilot, openai-codex
16 docker/metadata-action 250 22% claude-code, github-copilot, cursor, gemini-cli
17 github/codeql-action 216 42% claude-code, github-copilot, cursor, openai-codex
18 codecov/codecov-action 211 27% claude-code, github-copilot, cursor, gemini-cli
19 actions/configure-pages 205 18% claude-code, github-copilot, cursor, openai-codex
20 astral-sh/setup-uv 180 37% claude-code, github-copilot, cursor, openai-codex
21 dtolnay/rust-toolchain 175 21% claude-code, cursor, github-copilot, openai-codex
22 docker/setup-qemu-action 161 22% claude-code, github-copilot, cursor, openclaw
23 pypa/gh-action-pypi-publish 135 30% claude-code, cursor, github-copilot, gemini-cli
24 anthropics/claude-code-action 131 30% claude-code, github-copilot, cursor, gemini-cli
25 swatinem/rust-cache 128 26% claude-code, github-copilot, cursor, openai-codex
26 actions/setup-go 120 39% claude-code, github-copilot, cursor, openai-codex
27 actions/create-github-app-token 111 60% claude-code, github-copilot, cursor, openai-codex
28 actions/setup-java 104 25% claude-code, github-copilot, cursor, gemini-cli
29 actions/stale 103 40% claude-code, cursor, github-copilot, openai-codex
30 peter-evans/create-pull-request 93 55% claude-code, github-copilot, cursor, openai-codex
31 actions/dependency-review-action 91 46% claude-code, github-copilot, cursor, openai-codex
32 changesets/action 85 38% claude-code, github-copilot, cursor, openai-codex
33 dorny/paths-filter 83 45% claude-code, github-copilot, cursor, openai-codex
34 actions/setup-dotnet 82 21% github-copilot, claude-code, cursor, gemini-cli
35 github/codeql-action/init 78 53% claude-code, github-copilot, cursor, gemini-cli
36 github/codeql-action/analyze 78 53% claude-code, github-copilot, cursor, gemini-cli
37 aws-actions/configure-aws-credentials 70 41% claude-code, cursor, github-copilot, openai-codex
38 aquasecurity/trivy-action 70 54% claude-code, github-copilot, gemini-cli, openclaw
39 actions/attest-build-provenance 69 51% claude-code, github-copilot, cursor, openai-codex
40 actions/labeler 62 50% claude-code, github-copilot, cursor, openclaw

Before you add an Action: read its repository, give the workflow only the permissions: it needs, keep keys in repository secrets, and pin third-party Actions to a full-length commit SHA (GitHub's security guide). Facts above come from each Action's action.yml, latest release and repository, checked Nov 16, 2026. For AI assistants: agent.md and the MCP tool find_github_actions.