AG-UI
An event protocol for bringing agent backends into user-facing applications.
- Source repo
- ag-ui-protocol/ag-ui
- Stars
- ★ 17k
- Last updated
- 1d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 34/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformClaude API (Partial support)
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A team building a real-time chat interface that needs to carry an agent backend’s streaming execution events into a user-facing app.
- Not a fit if
- Teams wanting a turnkey agent that performs business tasks on its own
- Users without their own agent backend and frontend application
- Teams needing documented event names, endpoints, and auth before adopting
- Source review
- 34/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
AG-UI is an open, lightweight, event-based protocol for agent–user interaction, not a turnkey agent product. During execution, an agent backend emits events compatible with roughly 16 standard event types and accepts a small set of compatible inputs. Its middleware layer supports arbitrary event transports, including SSE, WebSockets, and webhooks, with loose event-format matching. The project ships a reference HTTP implementation and a default connector for wiring backend activity into user interfaces. The repository also lists framework, community SDK, and client integrations, but the supplied material does not document generated-app runtime, deployment, authentication, or credential setup.
During an agent run, an AG-UI-integrated backend produces a stream of events compatible with roughly 16 standard event types and accepts AG-UI-compatible inputs as arguments. Middleware adapts that interaction to SSE, WebSockets, webhooks, or another event transport, enabling the listed capabilities of real-time streaming, bidirectional state synchronization, structured messages, and generative UI. The project also includes a reference HTTP implementation and a default connector; npx create-ag-ui-app my-agent-app is the only explicit app-creation command in the supplied material. No event names, HTTP endpoints, payload schemas, environment variables, or production deployment configuration are provided.
- A team building a real-time chat interface that needs to carry an agent backend’s streaming execution events into a user-facing app.
- A product engineer who needs bidirectional state synchronization between a frontend and an agent execution.
- A LangGraph or CrewAI team seeking one of the AG-UI paths marked supported in the README.
- A frontend developer adding structured messages, generative UI, or frontend tool interaction to an application interface.
- An integration team that needs to use one interaction protocol across HTTP-style, SSE, WebSocket, or webhook transports.
How do you install or deploy this agent?
The supplied material provides this initialization command: npx create-ag-ui-app my-agent-app. It creates a new AG-UI application. It does not state a required Node.js version, post-generation start command, dependency-installation flow, credentials, environment variables, or deployment configuration, so a complete verified installation and deployment procedure cannot be derived from it.
How do you use this agent?
Run npx create-ag-ui-app my-agent-app to create an application, then configure an agent backend to emit AG-UI-compatible events during execution and accept AG-UI-compatible inputs. Connect it through the reference HTTP implementation or default connector, using SSE, WebSockets, webhooks, or another event transport. The supplied material does not include event definitions, connector call examples, or a generated-app start command, so a first runnable invocation cannot be verified from it alone.
What are this agent's strengths and limitations?
- It is explicitly centered on the interaction boundary between agent backends and user interfaces, rather than only on model calls.
- The README states that it works with arbitrary event transports and specifically names SSE, WebSockets, and webhooks.
- A reference HTTP implementation and default connector provide concrete starting points for protocol adoption.
- The repository lists supported or integrated paths for LangGraph, CrewAI, Google ADK, and Microsoft Agent Framework, among others.
- It is a protocol and integration layer, not an agent that performs business tasks on its own; adopters still need an agent backend and a frontend application.
- The supplied material does not specify event names, endpoints, authentication, or error-handling semantics, so implementation requires consulting the specification.
- Several listed integrations remain in progress or help wanted, including OpenAI Agent SDK, Cloudflare Agents, .NET, Nim, Flowise, Langflow, and React Native.
- Beyond the initialization command, the material does not document runtime versions, production deployment, or credential requirements.
How does this agent compare with similar options?
AG-UI is positioned as complementary to MCP and A2A: MCP gives agents tools, A2A enables agent-to-agent communication, and AG-UI brings agents into user-facing applications.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AG-UI This agent | 34 · Major gaps | CLIFree | ★ 17k | 1d ago | TypeScript | — |
| TrueForge | 67 · Some gaps | CLIFree + model costs | ★ 6.4k | 1d ago | TypeScript | OpenAI API · Claude API |
| Tambo AI: Generative UI Toolkit for React | 42 · Major gaps | Library / SDKFreemium | ★ 11k | 27d ago | TypeScript | OpenAI API · Claude API |
| Mirafold | 89 · Good | CLIFreemium | ★ 19 | 1mo ago | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows no specific implementation or documentation for permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. All criteria scored 0 due to lack of evidence.
Self-consistency scored 2: README, package.json, and test files are largely consistent in protocol description and integration lists. Dependency availability scored 1: package.json lists dependencies and overrides but no availability guarantees or lockfile details. Failure messages scored 1: tests include retry logic but no user-facing error message documentation.
Audience and scenarios scored 2: README clearly lists target audience (developers) and multiple use cases (chat, state sync, generative UI, etc.). Capability boundaries scored 1: protocol scope described but no explicit limitations. Trigger precision scored 1: event types described but no detailed trigger conditions. Environment fit scored 2: supports multiple frameworks and transports but no detailed configuration guides.
Information architecture scored 2: README well-structured with TOC and links. Install notes scored 2: provides npx create-ag-ui-app command. Naming stability scored 1: version 0.0.1, no stability promise. Examples and FAQ scored 2: provides examples and Dojo links. Known limitations scored 1: no explicit limitations listed. License scored 2: MIT license file present. Versioning and changelog scored 1: no CHANGELOG. Maintenance responsibility scored 1: author info present but no maintenance policy.
Output usability scored 2: protocol clearly defined with reference implementation. Marginal value scored 2: fills a gap in Agent-UI interaction protocols. Cost-benefit scored 1: no performance or cost data provided.
Claim traceability scored 1: README claims not linked to specific code or tests. Cross-source corroboration scored 1: multiple integration lists but not independently verified. Fact-inference separation scored 1: some content is inferred without clear distinction.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- No security-related documentation or implementation provided, such as permission management, data encryption, user confirmation; assess before use.
- Version 0.0.1 with no changelog; API may be unstable.
- Dependency security not audited; overrides present but reasons not explained.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
release/2026-11-05 Release 2026-11-05
Release notes excerpt
Packages Published Maven Central - published at 08:13:26 UTC | Package | Version | Install | |---------|---------|--------| | kotlin-core | 0.5.0 | com.ag-ui.community:kotlin-core:0.5.0 | | kotlin-client | 0.5.0 | com.ag-ui.community:kotlin-client:0.5.0 | | kotlin-tools | 0.5.0 | com.ag-ui.community:kotlin-tools:0.5.0 | [email protected] - Aligned the SDK with AG-UI 1.0, including strict event/input decoding and state patch structural checks. - Added REASONING_* events; THINKING_* events and ThinkingTelemetryState are deprecated with ReplaceWith hints. …View on GitHub ↗ -
release/2026-11-03 Release 2026-11-03
Release notes excerpt
Packages Published Python (PyPI) - published at 01:10:30 UTC | Package | Version | Install | |---------|---------|--------| | ag_ui_adk | 0.12.0 | pip install ag_ui_adk==0.12.0 | [email protected] - Added opt-in A2UI Express inference path alongside Direct JSON, selectable per run for injected and explicitly owned render tools. - Express output is compiled incrementally by a2ui-agent-sdk 0.10.0 into validated A2UI activity updates on a stable surface/activity identity. …View on GitHub ↗ -
release/2026-11-02 Release 2026-11-02
Release notes excerpt
Packages Published Python (PyPI) - published at 19:59:02 UTC | Package | Version | Install | |---------|---------|--------| | ag-ui-a2ui-toolkit | 0.0.6 | pip install ag-ui-a2ui-toolkit==0.0.6 | [email protected] - Added A2UIGenerationError. Adapters can raise it from the sub-agent call with structured errors; run_a2ui_generation_with_recovery records the failed attempt and feeds those errors back to the model on the next attempt, within the existing attempt limit. Other exceptions (transport failures, cancellation) still propagate unchanged. …View on GitHub ↗