Built for AI Agents

A command line and MCP server an AI agent can drive, with guardrails, approvals and locks you control.

BifurcSearch requests, mocks and environments…Ctrl PCommandsCtrl KDevelopment

Acme APIWorkspace

Applications

ServicesHealth bar

Routing

Mappings4Proxy rules3CaptureMail

API

Requests12Mocks5Flows1

Workspace

Environments2Vaults1ConfigAudit log

Settings

main2 unpublished changes on this screenPublish

Forward proxy :8080 runningEngine 0.6.0

How it works

bifurc agent init adds a short cheat sheet to AGENTS.md or CLAUDE.md so a coding agent knows how to send your saved requests, read captured traffic, wait for a service and check its mocks. Each agent can work in a named session with its own cookies and variables. A policy file in the workspace blocks calls, limits the hosts an agent may reach, or holds a call until you approve it in the app. An item someone is editing is locked against everyone else, so two agents, or an agent and you, never overwrite each other.

When to use it

• Let a coding agent run your saved requests to check its own change.
• Keep agents away from production with one deny rule.
• Approve or deny an agent's DELETE from a dialog in the app.