Agents Shipgate CLI
Agents Shipgate shows in a pull request how an agent permissions and tools changed before the changes reach main
- 87 ★ on the repository
- Last commit
- Apache-2.0
Minimal access
Why this level
- Reads repository configuration and the grants it declares
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agents-shipgateDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Reads repository configuration and the grants it declares
This is third-party code. Review the repository files before installing.
What it does
Agents Shipgate is a CLI and GitHub Action for static tool-use readiness review. It reads MCP, OpenAPI, OpenAI Agents SDK, Anthropic Messages API, Google ADK, LangChain and LangGraph, CrewAI, Codex configuration and plugins, n8n and Conductor OSS artifacts and writes a deterministic Tool-Use Readiness Report before an agent gets working permissions. It does not run agents, tool calls, an LLM or the network. The `agents-shipgate diff` command compares the working tree with the base branch and prints one row per changed grant: a widened Bash rule in `.claude/settings.json`, a removed denial, a new MCP server in `.mcp.json`. It needs no manifest, policy, saved baseline or account. For the OpenAI Agents SDK and Google ADK, `diff --application` compares tool bindings from source. Every change comes with a review question and a command that reproduces the comparison. The report gives no verdict on intent or actual access: it states only facts from repository declarations. A ready GitHub workflow example adds the same comparison as an advisory PR comment.
Who it is for. Teams where Claude Code, Codex and Cursor edit `.mcp.json` and permission settings and a reviewer needs to see what the agent can now do.
Good fit when
- An agent in a PR widened a Bash rule and added an MCP server, and you need a diff row for each permission
- Before an agent gets working permissions you need a readiness report for its tools
- An MCP package is attached through `npx` without an exact version and the review should notice it
- An advisory comment on every PR should show changes to agent capabilities
Not a fit when
- You need to know what the agent actually did at runtime: the report describes only what the files permit
- You need a merge block on intent and business rules: the tool raises reviewer questions and makes no decisions
- The agent is built on an SDK outside the list of supported artifacts
Example request
Compare the agent permission changes in this branch with main through Agents Shipgate and list the widened grantsLimitations
- The report is static: it shows what the files permit, not what the agent did, and it does not confirm intent.
- Setup-free comparison supports a limited set of files: Claude Code, Codex, Cursor and VS Code settings and MCP configurations.
- `diff --application` arrived in version 1.2.0, and an older install needs `pipx upgrade agents-shipgate`.
- Surfaces `diff` does not read are listed in `docs/host-boundary-support.md`.
How to disable. Remove the workflow from `.github/workflows` and uninstall the package with `pipx uninstall agents-shipgate`.
FAQ
Does diff need any setup?
No: a manifest, policy, saved baseline, skill and account are not required, and the command compares the working tree with the PR base.
Does the tool run agents?
No. The static mode runs no agents, tool calls or LLM and uses no network.
What does the report show about a new MCP server?
The command name, environment variable keys and that the `npx` package has no exact version, so the launch may resolve a local dependency or the registry.
Which artifacts does Agents Shipgate read?
MCP, OpenAPI, OpenAI Agents SDK, Anthropic Messages API, Google ADK, LangChain and LangGraph, CrewAI, Codex configuration and plugins, n8n and Conductor OSS.
How do I get the report on every merge request?
Add the example workflow `examples/github-actions/14-host-only-advisory-pr.yml`: it posts the same comparison as an advisory comment.
Discussion
No questions or reviews yet.
Related
CLI toolsAgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
A static scanner for vulnerabilities in AI agent code and MCP configs with 66 rules mapped to OWASP Agentic Top 10
Agentic Radar analyzes multi-agent code on LangGraph, CrewAI, n8n, OpenAI Agents and AutoGen and builds an HTML report on tools and vulnerabilities
A CLI for auditing agent MCP servers: inventory, evidence per server and a CI policy where unmeasured never counts as clean
ai-engineering installs hooks that deny risky agent calls and write a receipt for every verdict into your repository
Agentic SAST scanner: AI agents read your code, confirm findings through the call graph and scan a whole repository or a diff for pull request review
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Agents Shipgate (this card) | no | 87 | |
| AgentLint | no | 156 | a live guard that blocks secret writes and risky commands and does not show the permission delta in a PR |
| CC Safety Net | no | 2,014 | blocks risky Git and file system commands and does not compare agent permissions in a repository |
README in short
The README starts with the point: a deterministic check of agent capability changes, local and static. It then covers PR comparison without setup, sample output for widened grants, the five kinds of answer, advisory comment setup in GitHub and doc links.