™Delegate across coding agents.
Hand a task to Claude Code or Codex with the context it needs. Keep the handoff and returned work together.
AgentD · Developer Platform
Coordinate your
AI coding agents.
Delegate the coding. Keep the context. Review the work your agents return.
Illustrative workflow
Claude CodeLead agent
CodexCode Sol
Claude CodeReview Opus
Return to leadCode + review notes
One GatewayModel choice per agent
Runs in your development environment. Tested on Linux.
Explore the details
Choose your agents and models.
AgentD · Task coordinationDelegate work. Collect results.
Unified Gateway · Model accessAdapt protocols. Route model calls.
01 / Lead & delegated workers
Delegated by Claude Code · Sol
Code + review notes → Claude Code
02 / Model selection
03 / Inference endpoint
Illustrative workflow. AgentD coordinates the work; the Gateway routes model calls. Each agent’s model is configured independently. Cross-provider combinations require compatible adapters and end-to-end validation.
Built for your coding workflow
Better checks. Less review prep.
Multi-model Gateway.
Choose compatible models for Runtime-assisted features through MegaNova Gateway. Requires configuration; Gateway access and usage billing are separate.
Explore model options →Run checks before review.
Attach your project’s test command to a task. Add a separate validator when you want a fresh agent to check the work.
Review without log hunting.
See commits, check results and requests for input in one place. Spend less time gathering context for your review.
Pick up where you left off.
Keep decisions, constraints and unfinished work across sessions. Give the next agent the context it needs to continue.
Keep work within bounds.
Set a time budget for each task. Track elapsed time and available usage estimates alongside the result.
You choose the checks and decide what to merge.
Your next project
Build with your agents.
Open Developer Console for your LibraOS account. Run AgentD alongside your coding tools in your development environment.
Where does AgentD run?
AgentD runs in your development environment and coordinates your connected coding agents. The tested setup is Linux.
Do I need LibraOS Runtime?
Task tracking, handoffs, checks and the dashboard work without Runtime. Knowledge search and model-assisted features need a configured LibraOS Runtime.
Explore Runtime →How does model access work?
Bring your coding-tool accounts and compatible model endpoints. Coding-tool subscriptions and model usage are separate. MegaNova Gateway has its own access and usage billing.
Explore model options →