For AI agents

Agents welcome

Your AI assistant can research us, send Rashmi a message, ask to start a Thought partner plan, or request a custom proposal. It does this through our MCP server or REST API, against the same versioned offers people see on this site.

How agents work with Embark GTM Agents discover Embark through llms.txt, the OpenAPI document, the MCP server and structured data. They then send Rashmi a message, start a Thought partner plan through Stripe Checkout that the person pays, or request a proposal that Rashmi approves. Every step is written to an append-only audit log. 1. Discover llms.txt · openapi.json · /mcp (MCP server) · schema.org structured data · get_offers (versioned prices and terms) 2. Act Agent calls a toolEmbark systemHuman checkpointOutcome send_message a question, or an office hour Message inbox stored for Rashmi Rashmi replies by email she arranges any call herself Conversation started start_checkout Thought partner plan Stripe Checkout subscription, terms required Person reviews and pays agents can't complete payment Engagement starts request_proposal brief, budget, timeline Review queue brief stored for Rashmi Rashmi approves and sends nothing binds until signed Scoped proposal 3. Audit log Every action and outcome is appended to embark-audit with the offer version in force. Entries can't be edited or deleted.
Agent Embark system A person decides Result

The ground rules

Agents act for a named person. People confirm, pay and sign.

Agents can

  • Read everything: services, plans, prices, terms
  • Send a message, including asking for a free office hour
  • Start a checkout or request a proposal

A person must

  • Receive Rashmi's reply at their own email
  • Review the terms and pay on Stripe
  • Sign anything custom, after Rashmi approves it

We keep

  • An append-only log of every action, with the offer version in force
  • Personal details apart from the log, so they can be deleted
  • A rate limit of 20 actions a minute per caller

Connect

One set of actions, two ways in.

MCP server

Add as a remote MCP server (Streamable HTTP) in Claude, ChatGPT or any MCP client.

https://www.embarkgtm.com/mcp

REST API

Described by an OpenAPI document. Start with GET /api/offers.

curl https://www.embarkgtm.com/api/offers
# OpenAPI: https://www.embarkgtm.com/openapi.json
ToolRESTWhat it does
get_offers GET /api/offers Embark GTM's services, plans, prices, terms and the policy for agents. Read this first; it is versioned, and every checkout and proposal records the version it was made against.
start_checkout POST /api/checkout Creates a Stripe checkout link for the Thought partner plan ($2,500/month, two-month minimum). Agents cannot pay: give the link to the person, who reviews the terms and pays themself. Until online checkout opens this returns checkout_unavailable; use send_message instead.
request_proposal POST /api/proposals For project-based or fractional work. Submits a brief; Embark GTM reviews it and replies to the person with a proposal. Nothing binds until both sides sign.
send_message POST /api/messages Send Embark GTM a question or note, or ask for a free Office hour (a 30-minute call with Rashmi Vittal). Rashmi replies to the person by email and arranges any call herself. For scoped work use request_proposal.
get_proposal_status GET /api/proposals/{id} Status of a proposal request: pending_review, sent or declined.

Also for agents: /llms.txt (summary), /index.md (this site as Markdown), /openapi.json.