Agently

Turn model output into verifiable interfaces and multi-step agent behavior into orchestrated systems.

Join the WeChat communityJoin WeChatScan to request access

TIP

The official installable Agently Skills are now available. Install the official skills for Codex / Claude Code instead of using the old docs pack or zip. Guide: /en/agent-docs

TIP

For real integrations, the recommended runtime path is now Async First. If you are building services, streaming UI, or TriggerFlow workflows, start from /en/async-support in parallel with Quickstart.

Quickstart

Run model setup, structured output, and Instant in minutes.

Open

Async First

The recommended production path for services, streaming UI, and TriggerFlow.

Open

Model Settings

OpenAICompatible config and common model access.

Open

Workflow & Extensions

Tools, MCP, KB, FastAPI, and the point where orchestration begins.

Open

Output Control

Agently Output Format, ensure_keys, and Instant structured streaming

Open

Model Response

Unified access to text, structured data, metadata, and events

Open

Prompt Management

Layered prompts, quick syntax, config, and variable mapping

Open

TriggerFlow Orchestration

when‑emit signal loops for complex event systems

Open

Agent Extensions

Tools, MCP, auto_func, KeyWaiter, Session/Memo

Open

Agent Systems Playbook

Scenario → capability → operations with code and outputs

Open

Case Studies

From KB Dialog and Daily News Collector to Talk to Control, see how capabilities become systems

Open

Official Agently Skills

Install the official skills for Codex / Claude Code instead of the old docs pack

Open