AgentRun documentation
Set up the platform, create an agent, and manage its runs through one API and console. Choose your harness and compute, recover interrupted work, and reuse what your agent learns.
Documentation follows AgentRun release · Updated .
Platform guides
Agents and runs
Define an agent, choose its harness, and follow a run through completion.
Models and API keys
Configure model providers and credentials for agent runs and learning.
Self-improving agents
Turn completed work into reusable skills and validated workflow drafts.
Durability and compute
Choose where agents execute and recover work after an interruption.
Control plane
Manage definitions, runs, files, and learned artifacts from one place.
Use the API
Create agents, start runs, and inspect results through the common API.
Implementation reference
Detailed configuration, commands, and interfaces from the AgentRun repository.
Operations
Platform reference
Harnesses & computers
Client reference
Advanced workflows, integrations, and project reference
Workflow authoring, optional extensions, package details, and contributor documentation.