Claude Code for GTM in 2026

In 2026 the GTM engineer often starts in a coding agent. GTME Pulse's benchmark already has AI coding tools in most of those setups, next to a CRM and an enrichment layer. Claude Code is a natural place to describe a monitor: who to watch, what counts as a signal, where the rows should land.

Claude Code will do that job on a vanilla setup. You can also spend the week connecting it.

What vanilla Claude Code makes you own

A serious GTM workflow touches a lot of systems. Search and scraping want Apollo, Apify, or Bright Data. Email and phone want a waterfall. The CRM wants HubSpot. The alert wants Slack. The send wants Instantly, HeyReach, or Smartlead. Ads want LinkedIn, Google, or Meta. The monitor wants a clock, a place to store yesterday's rows, and a log when a run fails.

Each of those can be an MCP server, a CLI, or a script with a key in the environment. Claude Code will wire them. You then own auth for each one, rate limits for each one, and a host for the thing that must run at 8am whether or not the laptop is open. Monitoring workflows are the part that turns a good session into a side system. The session ends. The cron does not exist yet. The dedupe list is a JSON file in the repo. The keys are wherever the session left them.

That setup is reasonable for a one-off investigation. It is a poor production runtime for an agency running a different monitor per client.

Monial as the single GTM toolkit

Monial is the runtime those sessions deploy into. Workflows are TypeScript. They execute in a remote sandbox. The functions are already in scope:

• scrappers.* for posts, profiles, jobs, news, and pages, with Apollo, Apify, and Bright Data managed
• enrichments.getEmail and enrichments.getPhone through a FullEnrich waterfall
• integrations.* for HubSpot, Slack, Instantly, HeyReach, Smartlead, and ad audiences
• ai.generateOutput to structure scraper output into the schema you return

Lists persist between runs. A cron or an inbound webhook starts the file. Credits are spent while the run executes. Building the workflow and storing the list do not.

You can ignore Claude Code and use Monial's own agent: describe the workflow in the product and it writes the file. Chat support is there when you want a person to look at it. Claude Code is for teams that already live in the editor and want the same runtime without a second way of working.

Cursor uses the same MCP. The rest of this page says Claude Code because that is the session most of these prompts are written in. The tools are identical.

Connect the MCP

MCP uses OAuth. API keys are for the REST API only.

1. Open Console → Settings and copy the MCP server URL. It is your app origin plus /mcp.
2. In Claude, add a custom connector with that URL and finish the OAuth prompt.
3. In Cursor, the same URL goes in mcp.json:

{
  "mcpServers": {
    "monial": {
      "url": "https://www.monial.co/mcp"
    }
  }
}

The tool list and field reference live in the MCP docs.

How a session should call it

The tools are the same operations as the API. Call them in this order so the file matches the sandbox:

1. get_workflow_skills before writing anything. This returns how scrapers, enrichments, integrations, lists, AI, and webhooks are called, plus how credits are estimated.
2. get_scrappers for the function names and arguments. Then get_scrapper_type for each result you intend to read. Guessing fields is how a workflow returns empty objects.
3. get_integrations before an integration call. Pass a name (hubspot, slack, instantly, heyreach, smartlead, linkedinAds, metaAds, googleAds) to see one provider, including whether it is connected.
4. create_list if the monitor needs memory. Workflow code addresses a list by id.
5. create_workflow with the TypeScript body, an output schema, and a cronTab if it should run on a schedule. Omit the cron and the inbound webhook is the trigger.
6. estimate_workflow_cost before you burn a large run.
7. run_workflow, then get_workflow_run for logs and output. list_workflow_runs is the history.

list_workflows, get_workflow, and update_workflow are how you change a file that is already live. update_workflow leaves omitted fields alone, so a session can pause a monitor with isActive: false without rewriting it.

A prompt that matches the tools

Paste something this specific. Vague prompts produce workflows that call functions with guessed shapes.

Using the Monial MCP, build a daily monitor. Find LinkedIn jobs posted in the last 7 days for "RevOps" and "sales operations" in the United States, cap 25. Call get_workflow_skills, get_scrappers, and get_scrapper_type before you write code. Keep a list of companies already alerted. Enrich net-new companies' likely champion for a work email only if a LinkedIn URL or a first name, last name, and company domain is present. Post the new rows to Slack. Set a daily cron. Estimate the cost, then run it once and show me the logs.

A one-off uses the same tools without a cron. "Run this workflow with this input" is run_workflow. An agent outside Monial can also POST to the workflow webhook and get the validated JSON back. That is the pattern for an SDR copilot: the copilot sends the account, the sandbox does the scraping and scoring, the copilot reads the result.

What to leave in Claude Code

Claude Code remains the place for work that is not a Monial workflow.

Use it to draft the ICP, read a client's site, and decide which monitor is worth hosting. Use it to review the TypeScript Monial's agent wrote before you save it. Use it for a one-off research pass you will not run again. Use it when you need a provider Monial does not wrap. The sandbox will refuse filesystem and open-network code on purpose.

Use Monial when the output has to happen again tomorrow, with the same definition, without the session open. The file is the workflow. The MCP is how Claude Code edits and runs it. The sandbox is why you do not keep a second stack of MCPs alive just to scrape, enrich, and notify.

If you want the longer comparison with Clay, Cargo, Deepline, WhiteWhale, and Clearcue, it is one page: Monial vs others. For Tasklet specifically: Monial vs Tasklet.