Use AgentDoctor in a terminal or CI pipeline.

AgentDoctor scans a repository and reports whether AI coding agents have enough project-specific instructions, commands, boundaries, and automation evidence to work safely.

Install

Windows PowerShell:

irm https://raw.githubusercontent.com/youssefsz/agentdoctor/main/install.ps1 | iex

macOS and Linux:

curl -fsSL https://raw.githubusercontent.com/youssefsz/agentdoctor/main/install.sh | sh

From a local source checkout:

cargo install --path crates/agentdoctor-cli --locked --force

Quick start

agentdoctor
agentdoctor scan
agentdoctor scan --format json --no-interactive
agentdoctor init --dry-run --agents codex,claude,cursor

Running agentdoctor opens the terminal UI in an interactive terminal. In scripts, CI, pipes, or with --no-interactive, it falls back to the pretty scan report.

Commands

Command	Purpose
`agentdoctor scan [path]`	Scan a repository and print a pretty or JSON report.
`agentdoctor init --dry-run`	Preview deterministic setup files without writing them.
`agentdoctor config agents --set`	Save selected global agents for future scans.
`agentdoctor upgrade`	Install the latest GitHub release for the current platform.
`agentdoctor uninstall --yes`	Remove the current AgentDoctor executable.

Configuration

AgentDoctor uses CLI flags first, then project configuration from .agentdoctor.toml, then global user configuration, then safe defaults.

version = 1

[agents]
enabled = ["codex"]

[score]
minimum = 80

[commands]
format = "cargo fmt --all"
lint = "cargo clippy --workspace --all-targets -- -D warnings"
test = "cargo test --workspace"
build = "cargo build --workspace --release"

Findings and score

Scores are out of 100 across agent files, project-specific detail, commands, safety boundaries, repo hygiene, and automation/CI. Findings use stable IDs such as AD001 for a missing canonical AGENTS.md and AD015 for secret-like MCP config literals.

JSON reports include score evidence, facts, findings, recommendations, and flattened top-level evidence for automation.

Exit codes

Code	Meaning
`0`	Success.
`1`	Scan completed, but the score is below the configured minimum.
`2`	Usage error.
`3`	Configuration or engine error.
`4`	Unexpected error.

Full references

The complete CLI, configuration, and scoring references live in the open-source repository: github.com/youssefsz/agentdoctor/docs .