WCGW

An MCP server that gives AI assistants like Claude a fully interactive shell and file editing environment on the local machine, supporting concurrent background commands, syntax-checked code edits, project context loading, and three operational modes from read-only architect to full shell access.

Evaluated Mar 07, 2026 (0d ago) v5.6.1
Homepage ↗ Repo ↗ Developer Tools shell terminal bash zsh code-editing mcp local python file-operations coding-agent
⚙ Agent Friendliness
80
/ 100
Can an agent use this?
🔒 Security
74
/ 100
Is it safe for agents?
⚡ Reliability
70
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
80
Documentation
82
Error Messages
72
Auth Simplicity
85
Rate Limits
72

🔒 Security

TLS Enforcement
85
Auth Strength
72
Scope Granularity
65
Dep. Hygiene
78
Secret Handling
72

Shell/terminal execution MCP — highest risk category. Can execute arbitrary commands. CRITICAL: sandbox thoroughly, use allowlists, never give production access. High blast radius.

⚡ Reliability

Uptime/SLA
72
Version Stability
72
Breaking Changes
68
Error Recovery
70
AF Security Reliability

Best When

You want Claude to have genuine shell and editor access on your local machine to autonomously execute, test, and iterate on code with minimal interruptions.

Avoid When

You need sandboxed or remote execution, or you're uncomfortable granting an AI agent unrestricted shell access to your system.

Use Cases

  • Giving Claude Desktop or other MCP clients full shell and filesystem access for autonomous coding tasks
  • Running interactive terminal sessions with ANSI support as part of AI-driven workflows
  • Safe read-only code exploration in architect mode before switching to code-writer mode
  • Context-preserving task handoffs via saved project context files

Not For

  • Remote or sandboxed execution environments (runs directly on local machine with full access)
  • Teams needing audit logs or access controls over shell operations
  • Windows users not using WSL

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

OAuth: No Scopes: No

No authentication; runs locally with the permissions of the current user process.

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

Apache-2.0 licensed. Completely free; no external service dependencies.

Agent Metadata

Pagination
chunked
Idempotent
Partial
Retry Guidance
Not documented

Known Gotchas

  • Grants unrestricted shell access — agents can delete files, run network commands, etc.
  • Requires files be read before editing (protection mechanism, but adds tool call overhead)
  • Windows support only via WSL; native Windows shell not supported
  • Three operational modes (architect/code-writer/wcgw) must be chosen upfront per session

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for WCGW.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-07.

6470
Packages Evaluated
26150
Need Evaluation
173
Need Re-evaluation
Community Powered