Autonomous software work needs a control plane.
Phaedrus routes agent work, runs preflights, enforces quality gates, tracks outcomes, and produces audit evidence before code moves forward.
Not another coding assistant.
Phaedrus is the neutral operating layer above Codex, Claude Code, Cursor, Copilot, local models, repos, CI, and review flows.
Preflight first. Agent second.
Work enters from your system
A GitHub issue, Linear ticket, chat request, CI job, or internal app sends work into a stable run envelope.
Phaedrus checks readiness
Preflight verifies repo state, instructions, credentials, tests, branch posture, and execution risk before an agent starts.
Policy chooses the route
The run selects the agent surface, model, tools, autonomy level, and approval path without locking the caller to one provider.
Quality gates create evidence
Tests, checks, review passes, screenshots, logs, and failure notes become inspectable proof instead of mystery transcripts.
A receipt returns to the caller
The requesting system gets status, artifacts, cost signals, gate outcomes, patch references, and any human decision required.
Inspectable proof, not mystery transcripts.
Every meaningful run outcome carries its tests, checks, review passes, screenshots, logs, and failure notes — with the connector that produced each one named inline.
"run": "ph_8f3k2", "agent": "claude-code", "gates": { "preflight": "pass", "tests": "142/142" }, "artifacts": ["PR #311", "screens/checkout.png"], "cost": "$0.84", "human_approval": "slack:#releases"
Use the agents you already trust. Govern the work they do.
Phaedrus sits above coding agents, model providers, repos, and CI so teams can move faster without coupling their workflow to one vendor.
Agent surfaces
Wrap Codex, Claude Code, Cursor, Copilot, local models, and future agents without making any one of them the control layer.
Run environments
Support local, BYO machine, CI, and future managed execution while keeping credentials and workspace boundaries explicit.
Quality evidence
Attach tests, logs, screenshots, PR state, review output, and policy results to every meaningful run outcome.
Audit receipts
Return what happened, why it happened, what passed, what failed, what it cost, and where a human approved or vetted.