About us

Make building AI agents
effortless

aibuddy is our ready-to-use AI agent — perceive, decide, done. It's open source end to end: the runtime, tools, and patterns to ship agents that feel delightful, with unrivalled speed and control — all yours to read, fork, and run.

Explore the docs GitHub

What we keep

Principles we keep coming back to

  1. 01

    Execution should leave a trace

    observable traceable reviewable

    The execution loop, context compaction, tool calls, and state transitions should be recorded clearly for debugging, auditing, and review. Runtime behavior should not depend on implicit processes that cannot be observed.

  2. 02

    Core logic should not bind to concrete implementations

    interfaces dependency injection cross-platform

    The Agent core does not bind to a specific database, sandbox, model provider, or platform implementation. Concrete dependencies are injected at the composition layer, so web, desktop, and self-hosted deployments reuse the same core logic.

  3. 03

    Boundary errors should surface early

    TypeScript Zod data contracts

    Strict TypeScript, schema validation, and explicit data contracts constrain module boundaries. Problems that can be caught at compile time or input boundaries should not be deferred to runtime.

  4. 04

    Platform differences stay outside the runtime core

    adapters self-hostable MIT

    Web, desktop, and server environments share the Agent core. Differences in storage, authentication, sandboxing, and system capabilities belong in adapters instead of leaking into the runtime logic.

Maintainers

Core maintainers

aibuddy is maintained as an engineering system, not only as a code repository. Runtime design, documentation, examples, and community feedback are handled together so the project can keep its technical boundaries, release cadence, and tradeoff standards aligned.

J

Jianyang Zhang

AI Architect / Tech Lead · Independent AI Builder

Suzhou, China

With 13 years of experience in full-stack engineering, platform architecture, and technical delivery, Jianyang shifted his focus to AI engineering in 2023. He specializes in long-horizon agents: defining verifiable tasks, building controllable long-running execution systems, and improving them through evaluation loops. He independently designs, builds, and maintains aibuddy from product and architecture through multi-platform implementation and production delivery.

Long Horizon Harness Engineering Agent Evals
GitHub Resume Email

Get involved

Help shape what aibuddy becomes

Three ways to move the project forward — pick the one that fits the time and energy you have right now.

Give feedback

A sharp issue with a repro beats a silent star — it's how roadmap priorities actually move.

Docs gaps, API ergonomics, missing tools — if it got in your way, we want to hear about it.

Open an issue

Contribute code

aibuddy is built in the open, in a pnpm + Turborepo monorepo. Every layer is documented — jump into any of:

  • New tools — add a first-party agent tool or MCP integration.
  • New platforms — port the shared packages to a new runtime.
  • Docs & guides — improve architecture notes or translate pages.
  • Bug fixes — small PRs land fast.
See the repository

Sponsor the project

If aibuddy powers something you ship, consider sponsoring through GitHub Sponsors. It funds focused time on the runtime, the docs site, and community support.

Sponsors are credited in the README, and commercial sponsors get a logo slot on the home page.

Become a sponsor