Claude Code
Agentic coding tool with direct shell, filesystem, and git access in the user's local environment. Supports configurable approval modes.
Agent Passport
APAI.agent-passport.v0.1Agent Passport
local-toolunverified v0.1
Agent: Claude CodeVendor: AnthropicVersion: varies (rolling release)
Install modes:local-tool
↓
Can read
- ·Files and directories in current working directory (recursive)
- ·Git history and status
- ·Shell command output
- ·Environment variables exposed to the shell session
↑
Can write
- ·Files in current working directory
- ·Git commits and branches
- ·Shell-writable destinations the user's account can reach
↔
Can access
- ·Any network endpoint reachable from the local machine (via Bash tool)
- ·MCP servers configured in .claude/mcp.json
- ·Anthropic Claude API (for model inference)
$
Can spend
- ·Claude API tokens per session (billed to operator's Anthropic account)
○
Can expose
- ·File contents sent to Claude API for inference
- ·Shell output included in API context
✓
Will request approval for
- ·Tool calls in default permission mode (first-time or dangerous patterns)
- ·Destructive operations (rm, overwrite) in default mode
- ·External network requests via Bash in default mode
- ·Any action in plan-only mode (plan-first, then explicit approval)
v0.1 note: This passport is descriptive, not enforceable. Runtime enforcement is the Policy Pack + Gateway story (Phase 4+). Fields sourced from publicly documented agent behavior as of May 2026.
Compatible packages (v0.1)
APAI packages that declare support for local-tool environments compatible with this agent. v0.1 match is by platform overlap; full capability-field matching ships with Phase 4.