Module 01 · Get connected
Outcome: a verified MCP connection, the full tool list in front of you, and one free tool you call before and after anything expensive.
- Surface
- MCP server
- Level
- Beginner
- Uses
check_credits- Credits
- 0 — check_credits is free
- Prerequisite
- A Sync GTM account
Tools in this module
| Tool | Cost | Required input | Use it when |
|---|---|---|---|
check_credits | Free | None | Before a batch, and again after, to learn what it cost |
One tool. It is the only free one on the server, and it is the one people skip.
Set up
Install the server in your client
Setup differs per client. MCP server setup has a page each for Claude Code, the Claude app, Claude Cowork, Cursor, Codex, ChatGPT, Gemini, OpenClaw and Manus.
Base URL: https://api.syncgtm.com/mcp. Sign-in is through the browser — no API key to paste.
Verify with a free call
Use the SyncGTM check_credits tool and tell me my current balance.A number coming back means the connection, the auth and the tool routing all work.
List what you have
List the SyncGTM MCP tools you have access to, grouped by category.
Show the tool name only, not a description.You should see roughly 50 across person enrichment, person search, LinkedIn profiles, company enrichment, growth, jobs, post engagement, social, scraping and ads.
Run one real lookup
Use linkedin_profile_enrich on this profile and give me the current title
and company only: <paste a LinkedIn profile URL you already know>Check the answer against what you already know. You are testing the data, not the tool.
Example prompt with filters
check_credits takes no parameters, so the useful pattern is bracketing — a credit check on both sides of the work.
1. Run check_credits and tell me the balance.
2. Use find_people with current_titles ["VP Sales","Head of Sales"],
current_company_headcount_ranges ["51-200"], person_countries
["United States"], limit 10.
3. Run check_credits again and tell me exactly what step 2 cost.Do that once per new workflow. You learn the real per-row cost on a batch small enough not to matter.
Workflow prompts
A single tool call is rarely the job. These are the shapes you actually send — SyncGTM doing the data work, your other systems doing the rest.
Cost a workflow before you run it for real
- run a credit check
- run the workflow on 10 rows only
- run a second credit check
- extrapolate to the full list before committing
Use the SyncGTM MCP server.
1. check_credits — tell me the balance.
2. Take only the FIRST 10 rows of the list below. For each: find_work_email,
then verify_email. Skip anyone with no result — do not guess a pattern.
3. check_credits again.
4. Tell me: credits spent, cost per row, hit rate, and what the full
<N> rows would cost at that rate. Do not run the rest.
<paste 10 rows>Post a low-balance warning to Slack
- check the SyncGTM balance
- compare it against the run you are about to do
- post to Slack if it will not cover it
1. check_credits on SyncGTM.
2. I am about to enrich 400 contacts at roughly 1.3 credits each.
3. If the balance is under 600, post a message to the #revops Slack channel:
"SyncGTM balance is <X> credits — the 400-contact enrichment needs ~520.
Top up at https://app.syncgtm.com/settings/credits/buy before it runs."
4. If the balance covers it, say so and do nothing else.Best prompting practices
- Say “use the SyncGTM MCP server” in the first message of a session. It stops the model answering from memory.
- Name the tool —
check_credits, not “how many credits do I have”. Tool names are unambiguous; descriptions are not. - Bracket anything new. Credit check, small run, credit check.
- State volumes. “The first 20” is a request you can afford to get wrong. “Everyone in this list” is not.
Where this breaks
An answer that arrives instantly with no credit movement means no tool ran — the model answered from training data. Profile and company details will be confidently out of date. Re-prompt naming the tool. More fixes in Common issues.
A connection that worked yesterday and fails today is almost always an expired browser session, not a broken install. Re-authenticate in your client’s connector settings before reinstalling anything.
Further automation
Once a chain works, save it rather than retyping it. A GTM agent is a chain stored as a skill file, already scoped to the right tools, filters and caps, triggered with a slash command.
This module ships as MCP Quickstart — /mcp-quickstart verifies the connection, lists the tools and brackets a run with credit checks.
Next module
02 — Find people, the four ways to produce a list of profiles, and which one is cheapest for the shape of list you want.
Reference: MCP server setup · Check Credits · Tools · Prompting guide