Complete applications built on the Agents API guide, each running against a real Core installation with a Project API key.

Parsar Agent workbench

A conversation workspace built on Core. Assemble Agents from a resource library and an environment template, then talk to them in a local Web chat or through Feishu bots that act as the Agent's identity. Both entries create ordinary public Sessions.

Source: example/parsar.

Run and configure it with the example README, which also owns its supported features, storage and validation.

What to look at

Each feature maps to one part of the API. Read the code next to the guide section:

Feature	API used	Guide	Code
Server-side key and headers	Bearer key, OpenAI-Beta, route allowlist	Before you start	server.mjs
Agents with a Harness and model	x_agents_core.harness, model_provider	Core extensions	server/sessions.mjs
Web and Feishu conversations	Public Session creation, input events, Turns and Items	Send input	server/sessions.mjs, server/feishu.mjs
Hosted Environments from templates	openai_hosted with packages, files, env and setup_commands	Create a Session	server/runtime-profile.mjs
Crash-safe Session creation	Idempotency-Key, recovery from a lost response	Idempotency	server/sessions.mjs
Live replies	Event stream, text deltas, reconnect and reconcile	Stream events	src/lib/live-session.ts
Follow-ups and cancel	Input events with idempotency keys	Send input	src/Composer.tsx
History	Paging Turns and Items	Pagination	src/lib/api.ts
Skills and versions	/skills, default version	Skills	src/Skills.tsx

RL and evaluation rollouts

A Python script that runs batches of tasks as hosted Sessions, with any Harness and model, and grades each answer outside the sandbox. It shows Core as a rollout engine for RL training and evaluation. Each rollout is one Session with no network, task files as initial files and the answer as an Artifact.

Source: example/rl-eval.

Run and configure it with the example README, which also owns its output format, task format and training setup.

What to look at

Feature	API used	Guide	Code
One Session per rollout	Inline input, metadata and openai_hosted	Create a Session	rl_eval.py
Task files without network	Inline files and network.access: disabled	Initial files	rl_eval.py
Harness and model matrix	x_agents_core.harness, model, model_provider	Core extensions	rl_eval.py
Wait, time out and cancel	Turn list polling, agent.session.input.cancel	Cancel	rl_eval.py
Answer outside the sandbox	Artifacts under outputs/	Artifacts	rl_eval.py
Trajectories and usage	Items and Turn usage	Turns and Items	rl_eval.py

To contribute an example, follow the application example boundary.