For agents

Your playbook, in whichever AI you already use

ChatGPT, Claude, Claude Code, Codex, Hermes or your own code connect to Bundly over MCP and read your playbook exactly as you see it: the offer, the buyer, every asset, the launch plan, and what needs you today. Deciding and editing with your agent comes next.

https://bundly.ai/mcpOAuth sign-in, or an agent key from Settings → AI agents.

Connect a client

ChatGPT

Plus, Pro, Team or Enterprise with connectors.

  1. Settings → Connectors → Create (or Add custom connector).
  2. MCP server URL: https://bundly.ai/mcp. Leave authentication on OAuth.
  3. Click Create, then Connect: sign in to Bundly and approve what ChatGPT may do.
  4. In a chat, enable the Bundly connector and ask for your playbook.

Claude

claude.ai on any paid plan.

  1. Settings → Connectors → Add custom connector.
  2. Remote MCP server URL: https://bundly.ai/mcp.
  3. Connect: sign in to Bundly and approve what Claude may do.
  4. Turn the connector on in a chat and ask for your next actions.

Claude Code

The terminal, the desktop app or the IDE extension.

  1. Run the command below, then `/mcp` inside Claude Code to sign in and approve.
  2. Ask: "what needs me in my Bundly playbook today?"
claude mcp add --transport http bundly https://bundly.ai/mcp

Codex

The Codex CLI and IDE extension.

  1. Add the server to your Codex config, then sign in when Codex asks.
# ~/.codex/config.toml
[mcp_servers.bundly]
url = "https://bundly.ai/mcp"

Hermes and other bots

Anything that cannot open a browser.

  1. Create an agent key in Settings → AI agents (shown once).
  2. Add an MCP server with the URL below and the key as the Bearer token.
https://bundly.ai/mcp
Authorization: Bearer bdly_…

Your own code

Any language, plain HTTPS.

  1. Create an agent key in Settings → AI agents.
  2. Call https://bundly.ai/api/agent/v1; OpenAPI at https://bundly.ai/api/agent/v1/openapi.json, the rules at https://bundly.ai/api/agent/v1/skill.md.
curl -H "Authorization: Bearer bdly_…" https://bundly.ai/api/agent/v1/account

What an agent can do today

  • get_account · plan, credits, this agent’s scopes
  • list_playbooks · every playbook with its stage and what needs you
  • get_playbook · the live view: summary, Markdown, JSON or the coding-agent brief
  • get_unit · one piece by ref, with what it is built from and what uses it
  • search_playbook · find a piece by its words
  • list_decisions · open and decided, with Bundly’s pick and who decided
  • get_next_actions · what needs the owner, ranked
  • get_changes · who changed what since a revision
  • get_job · progress of a build, rebuild or page check

Next: decide and edit (with undo and a reason on every change), then build and check pages under a daily cap you set.

What stays yours

  • You approve each client and pick what it may do: read, edit, decide, spend. Change it any time; the next call obeys.
  • Agents see exactly what you see. On a free playbook they get the free preview and the unlock link, nothing more.
  • Every change an agent makes is logged with who, what, before and after, and can be undone.
  • Your account, billing and these settings are never reachable by an agent.

Agent access is included with every paid Bundly and with Pro.