Claude Code
Governance
Anthropic's Claude Code is the most capable agentic coding tool available. It writes code, runs bash commands, and accesses MCP tools autonomously. OnLeash adds the governance layer that makes it safe for production teams.
Why Claude Code Needs Governance
Claude Code is powerful because it has deep system access. That's also why it's dangerous without oversight.
Full terminal access by default
Claude Code runs bash commands directly. Without governance, it has the same permissions as the user who launched it.
No built-in policy engine
Claude Code has permission prompts but no persistent policies, no audit trail, no compliance evidence, and no kill switch.
Extended tool access via MCP
MCP servers give Claude Code access to databases, cloud APIs, and external services. Each tool is an additional attack surface.
No behavioral monitoring
No built-in mechanism to detect behavioral drift, anomalous patterns, or gradual escalation across sessions.
What OnLeash Governs in Claude Code
Every capability. Mapped risks. Deterministic governance.
Bash command execution
Arbitrary shell access — installs, deletions, network calls, privilege escalation
Command allowlists with parameter validation. Network-accessing commands require explicit approval. Kill switch halts all execution.
File read/write
Secret access, code tampering, configuration manipulation
Path-based policy rules. Sensitive paths require approval. Diff analysis flags suspicious edits.
MCP tool access
Extended capabilities via external tools — database, cloud APIs, third-party services
Tool enumeration. Per-agent tool restrictions. Unauthorized tool calls blocked. Parameter-level validation.
Multi-step task execution
Complex autonomous workflows without human checkpoints
Rate limiting, circuit breakers, approval gates at configurable risk thresholds. Token metering tracks cost.
Context window management
Large context may include sensitive data from the codebase
PII detection scans action payloads. Redaction prevents sensitive data from appearing in logs.
Integration via MCP
Claude Code natively supports MCP servers. OnLeash ships an MCP server that intercepts Claude Code's tool calls and routes them through the governance engine. Setup takes minutes, requires no code changes.
Install OnLeash MCP server
One command sets up the governance proxy for Claude Code.
Add to your Claude Code settings
Register OnLeash as an MCP server in your editor configuration.
Define policies
Set rules via the dashboard, CLI, or API.
Govern
Every Claude Code action is now governed.
Same Governance, Different Tool
Both Claude Code and Cursor are agentic coding tools. OnLeash integrates with both via MCP — same policy engine, same dashboard, same audit trails. Govern your entire fleet from one control plane.
See Cursor AI Governance →Govern Claude Code in Production
Free Developer tier. MCP-native integration. No code changes required.
Start Governing Claude Code