gtm-agent — Elias Stravik

Open source agent for GTM.

11★ Stars

0⑂ Forks

TSTypeScript

README.md

Build and run GTM workflows from Slack

GTM Agent lets non-technical teammates keep the organization's GTM context, ICPs, and personas, qualify prospects, and build, run, and schedule saved workflows from Slack, by running the open source GTM Skills as one Vercel Eve agent connected to one GTM workspace repository.

✓ 100% free and open source ✓ One-prompt Vercel deploy ✓ Git-backed workspace history

Make Slack the front door to your GTM workspace

A teammate mentions GTM Agent in a channel or messages it directly. The agent picks the right GTM skill, reads the connected workspace repository, and replies where the team can see it. Once it has replied in a thread, replies there need no further mention. Every change to the workspace is pushed to the repository as one plain-language commit, so what the team decided is always in the history.

Choose between rebuilding prompts, switching tools, wiring a generic chatbot — or deploying one GTM agent in Slack

	GTM Agent	Repeated prompts	Standalone templates	Generic AI chat
Deploys to Vercel from one prompt, Slack and workflow project included	✅	❌	❌	❌
Installs the GTM Skills at build time from a pinned release	✅	❌	❌	❌
Pins one workspace repository at deployment	✅	❌	❌	❌
Keeps GitHub, workflow, and database secrets out of the sandbox	✅	❌	❌	❌
Asks every question with options through Slack's native controls	✅	❌	❌	✅
Commits every workspace change to the repository with a plain-language message	✅	❌	❌	❌
Qualifies prospects from public web evidence only, with no paid data tools	✅	❌	❌	❌
Runs saved workflows on Vercel against your own Neon Postgres database	✅	❌	❌	❌
Posts workflow results and approval requests straight into a Slack channel	✅	❌	❌	❌

GTM Agent is a deliberately narrow Eve template: one Slack interface, one set of GTM skills, one workspace repository, and one path for durable changes.

Ask in the channel. Get the GTM answer and the next action.

📈 Do the GTM work where the request appears

Ask for an ICP, a persona, a team change, or a saved workflow in Slack. The agent replies in the same thread with what it changed and, for workflows, where to look: the diagram, the runs, and the data.

⚡ Qualify a prospect from the conversation

Paste a company or a person and ask "is this a fit". The agent checks it against the saved ICPs and personas using public web evidence and replies with the verdict and its reasons.

Enrich connections and followers

Ask "enrich my network". The agent asks for your source and enrichment services, then builds a workflow with people, their current companies, and links in both directions. It supports CSVs, existing tables, and retrieval services, with any compatible direct providers or aggregators. Up to five current experiences per person is the default; source retrieval and enrichment share one run budget.

💬 Let a running workflow ask the team

When a hosted workflow needs a decision, it posts to the Slack channel you chose. A reply in that thread approves or steers the run without anyone opening a dashboard.

Deploy the agent and ask the first GTM question in three steps

1️⃣

Say it to your coding agent
Install the GTM Skills, then "get gtm-agent running for Acme in Slack". With the GitHub and Vercel CLIs signed in, the agent deploys this template, the Slack app, the workspace repository, and a workflow project with its database.

2️⃣

Click Allow
One Slack install page: choose the workspace, add the listed events and scopes, Allow. Then invite the bot to your GTM channels.

3️⃣

Ask in Slack
Mention the agent: "set up our GTM workspace". Its first save becomes the workspace on main; workflows run on the connected project from the first one.

Choose how to get started

Self-serve

For GTM teams working in Slack

Free

✓ One open source Eve Slack agent
✓ Every GTM skill, pinned to a release
✓ One git-backed GTM workspace repository
✓ Organization, member, ICP, and persona lifecycles
✓ In-conversation prospect qualification
✓ Saved workflows on Vercel with your own Neon Postgres database
✓ Workflow approvals and results in Slack

Done-with-you

Hands-on setup and rollout for your GTM team

Let's talk

✓ Everything in self-serve
✓ GTM Agent deployment
✓ Slack and GitHub setup
✓ GTM workspace repository configuration
✓ ICP, persona, and workflow design
✓ Workflow project, Neon, and AI Gateway configuration
✓ Team rollout, training, and best practices
✓ Ongoing maintenance and upgrades
✓ Dedicated Slack channel support

Deploy GTM Agent

Book a demo

Get your questions answered

What does GTM Agent deploy?

One Eve agent on Vercel with Slack as its only channel, the GTM Skills installed at build time from a pinned release, and a connection to one GTM workspace repository.

Do I have to build the workspace before deploying?

No. The deploy creates an empty private repository named gtm-<slug> and points the agent at it; say "set up our GTM workspace" in Slack and the agent's first save becomes the workspace on main. The repository name, minus a leading gtm-, is the workspace slug.

Where do the agent's changes go?

To the workspace repository, as commits on main authored as the owner of the GitHub token you provide. The agent's own computer lasts one conversation; only what is pushed survives.

How are secrets handled?

They never enter the sandbox. The sandbox firewall adds the GitHub token, only to git requests for the workspace repository, and the workflow project's bypass secret to requests on the way out. Hosted data is read through the workflow project's query route, and a CSV import is a run started with the file's rows, so no database credential exists on the agent. Model access runs through Vercel's AI Gateway with each project's own identity, no key anywhere.

Can the agent run GTM workflows?

Yes, on the second Vercel project the one-prompt deploy connects to the workspace repository's workflows/ folder with a Neon Postgres database. The agent builds workflows in its sandbox, pushes them, and starts, approves, and cancels runs on that project. Results and approval requests post to the Slack channel the workflow names.

Does the agent use paid data providers to qualify prospects?

No. Fit checks from Slack use public web evidence only. Paid providers are available to saved workflows through keys you set on the workflow project; the agent learns their names and never holds their values.

Which model does it use?

openai/gpt-6-luna-fast through the AI Gateway, or the model you name in GTM_AGENT_MODEL, with reasoning effort from GTM_AGENT_REASONING (high when unset). Both are read at build time.

What does it cost?

GTM Agent is free, open source, and MIT licensed. Vercel, Slack, GitHub, Neon, and model usage may be subject to their own plans and charges.

Put the next GTM decision in Slack

Your team asks in the conversation. GTM Agent brings the workspace, the skill, and the next action into the same thread. You click Deploy once.

✓ 100% free and open source ✓ One-prompt Vercel deploy ✓ Git-backed workspace history

{
  "request": {
    "method": "GET",
    "url": "https://api.eliasstravik.com/projects/gtm-agent"
  },
  "response": {
    "status": "200 OK",
    "headers": {
      "content-type": "application/json; charset=utf-8",
      "x-powered-by": "one person + systems"
    },
    "body": {
      "name": "gtm-agent",
      "description": "Open source agent for GTM",
      "language": "TypeScript",
      "stars": 11,
      "forks": 0,
      "repo": "https://github.com/eliasstravik/gtm-agent",
      "readme": "https://github.com/eliasstravik/gtm-agent#readme"
    }
  }
}