F+Foldkit Plus
HomeLearnPackagesComponentsDemosGitHub ↗

Agent

Guides

foldkit-agent

  • Overview
  • Usage
  • API
  • Why the API looks like this
  • See also

foldkit-agent-mcp

  • Overview
  • Sixty seconds: serve a bound runtime
  • Protocol errors vs application refusals
  • Streamable HTTP
  • HTTP identity and sessions
  • Diagnostics

foldkit-agent-webmcp

  • Overview
  • Sixty seconds: publish a bound runtime
  • What a capability becomes
  • Tools follow the Model
  • Forms as declarative tools
  • Registration lifecycle

foldkit-agent-a2a

  • Overview
  • Sixty seconds: card + task handler
  • Tasks and completion

foldkit-agent-native

  • Overview
  • Sixty seconds: generate actions
  • What a capability becomes
  • Runtime and identity
  • Schema bridge
  • Compatibility proof
  • Limits

Expose, do not reimplement

  • Overview
  • The mental model
  • Why have a contract at all?
  • A small end-to-end example
  • What an agent may see: context and resources
  • What an agent may do: expose Messages
  • Dispatch is not always completion
  • Availability and authorization are different questions
  • Bind once, serve several protocols
  • How this relates to Surface
  • When not to use this
  • Choosing what to read next

API

foldkit-agent

  • foldkit-agent

foldkit-agent-mcp

  • foldkit-agent-mcp

foldkit-agent-webmcp

  • foldkit-agent-webmcp

foldkit-agent-a2a

  • foldkit-agent-a2a

foldkit-agent-native

  • foldkit-agent-native

Open source, MIT
Foldkit 0.168 · Effect 4

Agent / foldkit-agent-mcp

AgentMcpDiagnostic

type · packages/agent-mcp/src/diagnostics.ts:46

type AgentMcpDiagnostic =
  | AuthenticateDefect
  | CapabilityDefect
  | RequestDefect
  | ToolListDefect
  | TransportEndDefect

Every diagnostic foldkit-agent-mcp reports.

Previousfoldkit-agent-mcp

Edit on GitHub ↗

This site is a Foldkit Plus application, rendered at build time with foldkit-ssr. Its source

FoldkitEffectGitHub