Adapters Overview

Adapters are the bridge between Baton's orchestration layer and agent runtimes. Each adapter knows how to invoke a specific type of AI agent and capture its results.

How Adapters Work

When a heartbeat fires, Baton:

1. Looks up the agent's adapterType and adapterConfig
2. Calls the adapter's execute() function with the execution context
3. The adapter spawns or calls the agent runtime
4. The adapter captures stdout, parses usage/cost data, and returns a structured result

Built-in Adapters

Local Coding Adapters

Adapter	Type Key	Description
Claude Local	claude_local	Runs Claude Code CLI locally
Codex Local	codex_local	Runs OpenAI Codex CLI locally
Cursor Local	cursor	Runs Cursor CLI locally
Gemini Local	gemini_local	Runs the Gemini CLI locally
OpenCode Local	opencode_local	Runs OpenCode locally
Pi Local	pi_local	Runs the Pi coding agent locally

Infrastructure Adapters

Adapter	Type Key	Description
Process	process	Executes arbitrary shell commands
HTTP	http	Sends webhooks to external agents

Adapter Architecture

Each adapter is a package with three modules:

packages/adapters/<name>/
  src/
    index.ts            # Shared metadata (type, label, models)
    server/
      execute.ts        # Core execution logic
      parse.ts          # Output parsing
      test.ts           # Environment diagnostics
    ui/
      parse-stdout.ts   # Stdout -> transcript entries for run viewer
      build-config.ts   # Form values -> adapterConfig JSON
    cli/
      format-event.ts   # Terminal output for `baton run --watch`

Three registries consume these modules:

Registry	What it does
Server	Executes agents, captures results
UI	Renders run transcripts, provides config forms
CLI	Formats terminal output for live watching

Shared Adapter Utilities

Several local adapters also use shared helper utilities from packages/adapter-utils:

• Session compaction — rotate or reset resumable sessions when run count, token volume, or age thresholds are exceeded
• Log redaction — scrub sensitive home-directory paths from logs
• Billing inference — normalize usage and cost reporting across providers

Prompt Composition

Supported local adapters can receive supplementary project instructions composed by Baton at heartbeat time.

That composed layer can include:

• project backstory
• compact context or full conventions
• critical governance reminders

Choosing an Adapter

• Need a coding agent? Use claude_local, codex_local, cursor, gemini_local, opencode_local, or pi_local
• Need to run a script or command? Use process
• Need to call an external service? Use http
• Need something custom? Create your own adapter