Quick Guides
Twelve things you can do with AgentField that no other agent framework ships out of the box.
Twelve capabilities that come built into AgentField. Each guide is a single composed snippet — copy it, run it, see the wow.
Trigger agents from GitHub, Stripe, Slack, or any custom webhook
Bind a reasoner to an external event source — signed and replay-safe.
Agent discovery — MCP-style for agents calling agents
Live registry of agents the LLM can pick from at runtime.
Trigger agents on memory changes
One agent writes, another agent's reasoner fires — pattern-matched, cross-scope.
Harness orchestration
Dispatch Claude Code, Codex, Gemini, or OpenCode as typed workers with cost caps.
Multi-step human approval that survives restarts
Pause an agent for a human, crash the server, resume where it left off.
A/B test agent deployments
Run two versions of an agent side by side and split traffic between them.
Access control between agents
Allow or deny cross-agent calls down to the function and argument level.
Replace API keys with agent identity
Sign outbound calls with your agent's DID — partners verify with a public key.
Cryptographic receipts for every AI decision
Every reasoner produces a signed, offline-verifiable record of what ran.
Signed audit chain for any workflow
Every step in a multi-agent workflow becomes a signed credential, exportable.
Multimodal agents — vision, image, and audio in one pipeline
One agent that takes images in and generates images, audio, or video out.
Realtime voice sessions through the control plane
A live browser voice call in one decorator — provider key stays in the control plane, every turn lands in the workflow DAG.
Production capabilities
The backend capabilities AgentField gives agent systems out of the box.
Trigger agents from GitHub, Stripe, Slack, or any custom webhook
Bind a reasoner to an external event source. The control plane verifies the signature, drops replays, and dispatches the event to your reasoner.