GPTFree
Agent

Agent

Open-source TypeScript framework for agents that ship with a UI.

Define a capability once, use it in agent, UI and API.

Intelligent agentDevelopment platform AgentFrameworkOpen sourceTypeScriptLLMUIAutomation
Visit official site →

Highlights

defineAction declares a capability once and reuses it across agent, React UI, HTTP API and integrations.
Embedded agent chat so users can delegate tasks and review results in the same interface.
Shared application state: the agent sees what the user is viewing, selecting and editing.
Skills, memory, automation and agent teams for long-running and complex work.
Built-in authentication, organisation management and fine-grained sharing permissions.

What you can do with it

  • Build a design tool where the agent produces UI that follows your design system.
  • Build a slide generator that creates branded decks from input or existing documents.
  • Build an analytics dashboard where users query data in natural language and get charts.
  • Compose several specialised agents to split a larger task across a team.

How to use it

  1. 1 Create a project with the CLI by running npx agent-native@latest create my-app.
  2. 2 Install dependencies with npm install and start the dev server with npm run dev.
  3. 3 Declare capabilities with defineAction, specifying inputs, outputs and logic.
  4. 4 Call the same capability from the front end so the React interface is generated.
  5. 5 Configure a custom model or the default, and connect PostgreSQL.
  6. 6 Add skills, memory and automation, then deploy and configure auth and permissions.

Who it is for

TypeScript developers building agent products with an interface.Teams that want self-hosted control over data and models.Engineers prototyping agent features without writing plumbing.

Pros

  • One definition drives agent, UI, API and integrations, which removes duplicate code.
  • Shared state makes agent assistance context aware rather than blind.
  • Fully open source and free, with self-hosting supported.

Limitations

  • Self-hosting means you own Postgres, secrets, model keys, auth and deployment.
  • Giving agents real write access carries risk until flows are proven.
  • The project may change APIs between versions, so pin releases and read the changelog.

What Agent is

Agent, or Agent-Native, is an open-source TypeScript framework for building autonomous agents that ship with a real interface. The core idea is that you define a capability once with defineAction, and that single definition becomes available to the agent, to a React UI, to an HTTP client and to external integrations, so the same logic is not written four times. The framework includes shared application state, so the agent can see what the user is looking at, an embedded chat for delegating tasks, skills and memory for longer context, scheduled or event-driven automation, and teams of specialised agents that split a complex task. It uses PostgreSQL for shared data, supports proprietary models as well as defaults, and comes with authentication, organisation management and fine-grained sharing, editing and management permissions. Example apps such as design, slides and analytics show the patterns. It is fully open source and free, with self-hosting encouraged.

What you can do with it

- Declare a capability once and reuse it across agent, UI and API.

- Embed a chat panel where users delegate tasks and review output.

- Give agents skills, memory and scheduled triggers for unattended work.

- Compose several agents into a team for a complex task.

Who it is for

- TypeScript developers building agent products with an interface.

- Teams that want self-hosted control over data and models.

- Engineers prototyping agent features without writing plumbing.

What to watch out for

You host it, so you own the operational load: Postgres, secrets, model keys, auth and deployment are all yours, and the built-in permission model still has to be configured correctly before real users touch it. Letting an agent take real actions means real risk, so keep writes behind review until you trust the flows, and log everything. The project is young by open-source standards, so interfaces may shift between versions; pin releases and read the changelog. Free and open source does not mean supported, so check the licence and the maintenance activity before committing production work to it.

FAQ

Is it free?

Yes. It is open source and free, and the project encourages self-hosting.

What language is it written in?

TypeScript, with React on the interface side and PostgreSQL for shared data.

How do I start a project?

Run the CLI command to create an app, install dependencies, then declare capabilities with defineAction.

Can I use my own model?

Yes. You can configure a proprietary model or use the default one.

Why it is free

What the official page says: 免费

Free limits, trial length and features are decided by the official site and can change at any time.

Agent

Build an open-source framework for autonomous agents and intuitive UIs, defining capabilities once for multi-end reuse.

Visit official site →

Related free tools