How-to guides

Each how-to here solves one concrete problem. They assume you already understand the basics (read the Tutorials first) and they don’t try to teach concepts (see Explanation for that).

Building agents

• Equip an agent with tools — pick capabilities and assign them.
• Give an agent web access — web_fetch, network policies, allowlists.
• Define agents as files — version-controllable agent definitions in Markdown, TOML, or YAML.
• Use AGENTS.md for project instructions — inject project-level context into the system prompt.
• Customize a harness — create your own harness as a starting point for many agents.
• Migrate between LLM providers — swap OpenAI ↔ Anthropic ↔ Gemini without rewriting agents.

Running agents

• Stream events with the SDK — consume the SSE stream from Python, with reconnection and event filtering.
• Consume events via raw SSE — when you don’t want the SDK: curl, EventSource, or any HTTP client.
• Handle errors and cancel turns — graceful failure paths, turn cancellation, retries.
• Orchestrate multi-agent pipelines — chain sessions together.

Packaging and distribution

• Package an agent skill — author a SKILL.md, bundle scripts and references.
• Publish a skill to the registry — share skills across agents.
• Publish an agent as a Slack app — deploy an agent to a Slack workspace.

Operating

• Automate with the CLI — scripting against the CLI with jq.
• Deploy with Docker Compose — bring up the full platform.
• Enforce a budget — cap token spend per agent, session, or organization.