Documentation

Documentation

Every Ghosty Studio agent is its own machine: disk, terminal and memory. Here is how to create one, where to put it to work, and how to talk to it from your software.

Which path?

Full index

Getting started
  • Getting started
    Create an agent, hand it a PDF and watch it download it to its machine, read it and answer with the exact figures. Ten minutes, no code.
  • Concepts
    Agent, machine, conversation, turn, channel, connector, skill, engine and plan — the vocabulary of Ghosty Studio.
  • Plan and progress
    On long jobs the agent shows its live plan, delivers each milestone in the thread and leaves the final answer below it.
  • Configure Ghosty from your coding agent
    Identity, model, knowledge files, skills and MCP servers — all through an API with the agent's token, plus a skill so Claude Code, Cursor or Codex do it for you.
  • Recipes
    An agent fits in a YAML file — title, description, instructions and channels. The eight house recipes, how to upload yours and how to export an agent's.
Channels
  • WhatsApp
    The number is connected in Sales (WhatsApp Business) and on the agent card you pick which numbers it answers; voice notes, images and PDFs.
  • Teams
    Activate a Studio agent in a Ghosty Teams workspace with an @handle; channels, DMs, documents and tasks.
  • Messenger
    Connect your Facebook page so the agent answers on Messenger and on your ads' comments, and teach Meta which leads actually buy.
Connectors
  • Connectors
    External accounts the agent accesses on your behalf — Shopify, Mercado Libre, Google Drive, Gmail, Stripe, Mercado Pago, Notion, HubSpot, monday.com, Zapier, Odoo, GitHub and more — and how to connect them from the web, the app or the CLI.
  • Browser (Ghosty for Chrome)
    The Ghosty for Chrome extension lends your browser to your agent, with the sessions you already have open. Install, use from /c, the app and the terminal, browser_* tools, shortcuts, scheduled tasks and privacy.
Software Factory
  • Software Factory — PRs you review in two minutes
    Three agents (@plan, @build, @check) work your repo from Ghosty Teams chat; you sign the plan and the PR. How to install it, how to ask and what every PR brings.
  • Team per repo and per request
    Which agent and model plays each Software Factory role — the workspace team, `.ghosty/factory.md` in the repo and "@build con opus" in the message.
  • Repo knowledge base
    What agents must know about your project lives in your repo — notes in `docs/agents/` with `AGENTS.md` as the index — and the Software Factory reads it and keeps it up to date through PRs.
  • Evals — pick agents and models with your own requests
    Re-run one role of an already merged request with another agent or model and have a judge score it against what was actually done. Rubrics, @eval, cost and limits.
  • Stuck runs and why a run failed
    How the Software Factory warns when a role gets stuck, what counts as stuck, and how to read why a run went wrong.
  • Your workspace's CI box
    What the machine that runs your repos' CI comes with, how to request it in a workflow and when to use the large one (8g).
Protocols
  • ACP — connect your editor
    Connect Zed, VS Code, JetBrains or neovim to a Ghosty Studio agent over the Agent Client Protocol.
  • A2A — agent to agent
    Every agent publishes an A2A v1.0 Agent Card so other agents can discover it; the message endpoint opens to third parties soon.
  • MCP — the agent's tools
    What tools a Ghosty agent has, how they are named and how they reach each engine through the Model Context Protocol.
  • Documentation MCP
    Install the Ghosty Studio documentation as an MCP server in Claude Code, Cursor or Codex; your agent searches and reads it without leaving the editor.
CLI
  • CLI — Ghosty from your terminal
    Install the official Ghosty Studio CLI, sign in with your account, and configure, test and chat with your agents from the terminal.
  • Command reference
    Every Ghosty CLI command with examples (agents, channels, connectors, credentials, turns, sites, browser), global options and exit codes.
  • CLI for coding agents
    How Claude Code, Codex or Cursor use the Ghosty CLI for you — agent-driven login, JSON output, exit codes and the ghosty-agent skill.
  • Partners — an agent for each of your customers
    With ghosty partner you give an assistant to every business that uses your product — one shared agent that only uses your MCP, memory per customer, and credentials for your server.
API
  • Authentication
    OAuth2 with PKCE to talk to the Ghosty Studio API on behalf of a person — scopes, tokens, errors and how to get a client_id.
  • API reference
    Every /api/v2 endpoint — agent configuration, databases, documents, turns, WhatsApp, boards, account, conversations, files and partners — with its OpenAPI 3.1 schema.
  • Events (SSE)
    Catalog of a conversation's stream events — started, chunk, tool, permission, artifact, done — and how to write a client that doesn't miss anything.
  • Errors and limits
    Response codes of the Ghosty Studio API, what each one means and how to retry; sizes and caps.
Reference
  • Changelog
    What changed in the platform and in this documentation, by date.
  • Security
    What isolates one agent from another, where your keys live, what an agent can do on a public channel and how an account is deleted.