Developer Documentation

Refine your agents. Ship with certainty.

Verify agent resilience with Pipelines Odyssey. Observe agent behavior in production-emulated environments built from your data. Validate performance before shipping.

Start Testing API Reference

Press K to search docs

Connect

Register an agent

Point Pipelines at your HTTP endpoint or pasted Python.

Agent SDK

Wrap any Python agent in ~15 lines and route tools through the proxy.

Tools schema

Declare each tool — simulate it or pass it through to a live endpoint.

Test

Task seeding

Define scenarios: instruction, behavior, world state, failure rules.

Multi-turn testing

Drive the agent with a model-as-user to test memory and consistency.

Inspecting runs

Read the trajectory, ledger diffs, and judge verdict for every run.

Build with the API

API Reference

Programmatic access to agents, runs, and datasets.

LLM providers

Bring-your-own-key models for your agents and the judge.

Local development

Bridge your localhost agent to the platform with one command.