GTM Engineer Quickstart - AmpUp Documentation

AmpUp gives a GTM engineer one governed context layer for meetings, transcripts, accounts, contacts, deals, analysis scores, coaching feedback, tasks, briefs, prospecting, and CRM workflows. A human owner creates the workspace and authorizes access. The agent then uses one of two interfaces:

Interface	Best for	Authentication
MCP server	Interactive agents that discover and compose tools	OAuth 2.1 (recommended) or API key
Direct API	ETL, scheduled jobs, and deterministic services	Revocable API key

Start a free workspace

Begin a 14-day trial, choose GTM Engineer during onboarding, and connect the systems the agent should understand.

Open the MCP reference

Connect Claude, Codex, Cursor, or another MCP-compatible agent with OAuth.

1. Create and authorize the workspace

Start the AmpUp workspace

Open app.ampup.ai/trial. The workspace owner completes signup, chooses GTM Engineer, and connects the relevant CRM, calendar, conversation-intelligence, and messaging integrations.

Choose the agent interface

Use MCP for an interactive agent. Use the direct API for a service or data pipeline whose calls should be explicit and repeatable.

Authorize only what the workflow needs

OAuth connections run as the signed-in user. API keys inherit the role and permissions of the user who creates them and can be revoked from Settings → API Keys.

A human should retain ownership of signup, billing, consent, and material side effects. Keep CRM sync, campaign launch, and outbound messages behind an explicit approval step.

2. Connect an interactive agent with MCP

The hosted endpoint is:

https://app.ampup.ai/mcp

For Claude Code:

claude mcp add --transport http ampup https://app.ampup.ai/mcp

For Codex:

codex mcp add ampup -- npx -y mcp-remote https://app.ampup.ai/mcp

Complete the OAuth sign-in when the client prompts you. The agent can discover the allowed AmpUp tools and their current input schemas through MCP tools/list; treat that response as the source of truth for the connection. See MCP Server for every client setup and the current tool surface, and MCP Server Authentication for token, scope, and tenant-isolation details.

3. Call the API directly

Direct calls use this base URL:

https://app.ampup.ai/mcp/api

The workflow endpoints are JSON POST operations. Store the key in a secret manager or environment variable; never put it in browser code or source control.

export AMPUP_API_KEY="sk-a79-..."
export AMPUP_API_BASE="https://app.ampup.ai/mcp/api"

List analyzed meetings

curl -sS "$AMPUP_API_BASE/v1/list_meetings" \
  -H "Authorization: Bearer $AMPUP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "status": "analyzed",
    "limit": 25,
    "offset": 0
  }'

Useful filters include account_id, opportunity_id, search, status, scheduled_from, scheduled_to, limit, and offset.

Get one meeting and its analysis

Start here before pulling a full transcript. The meeting overview includes core metadata, inline analysis when available, and available_actions flags that tell the agent which deeper artifacts exist.

curl -sS "$AMPUP_API_BASE/v1/get_meeting" \
  -H "Authorization: Bearer $AMPUP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"meeting_id":"MEETING_ID"}'

Bulk-pull analysis for a warehouse job

Fetch scores, feedback, key moments, and deal intelligence for up to 100 meeting IDs in one request:

curl -sS "$AMPUP_API_BASE/v1/list_meeting_analyses" \
  -H "Authorization: Bearer $AMPUP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_ids": ["MEETING_ID_1", "MEETING_ID_2"],
    "include_deal_intelligence": true,
    "include_feedback": true
  }'

Pull the full transcript only when needed

curl -sS "$AMPUP_API_BASE/v1/get_meeting_transcript" \
  -H "Authorization: Bearer $AMPUP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"meeting_id":"MEETING_ID"}'

Full transcripts can be large. For a quote or a specific section, prefer /v1/get_transcript_excerpt with meeting_id plus a start_time / end_time range or a chapter name.

Pull deal-level conversation intelligence

curl -sS "$AMPUP_API_BASE/v1/get_opportunity_analysis" \
  -H "Authorization: Bearer $AMPUP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"opportunity_id":"DEAL_ID"}'

This returns the deal metadata plus its linked meetings and inline analysis, which is useful for deal health, qualification, forecasting, and coaching.

Server-side JavaScript helper

Use a thin wrapper so every call has the same authentication and error handling:

const baseUrl = "https://app.ampup.ai/mcp/api";
const apiKey = process.env.AMPUP_API_KEY;

async function ampup<T>(path: string, body: unknown): Promise<T> {
  if (!apiKey) throw new Error("AMPUP_API_KEY is not configured");

  const response = await fetch(`${baseUrl}${path}`, {
    method: "POST",
    headers: {
      Authorization: `Bearer ${apiKey}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify(body),
  });

  if (!response.ok) {
    const detail = await response.text();
    throw new Error(`AmpUp ${response.status}: ${detail}`);
  }

  return response.json() as Promise<T>;
}

const meetings = await ampup("/v1/list_meetings", {
  status: "analyzed",
  limit: 25,
  offset: 0,
});

Run direct calls on a trusted server, worker, or job runner. Do not expose the API key in client-side JavaScript.

Python SDK equivalent

The SDK wraps the same workflow API:

pip install ampup

from ampup.client import AmpUpClient
from ampup.models.list_meeting_analyses_request import ListMeetingAnalysesRequest
from ampup.models.list_meetings_request import ListMeetingsRequest

with AmpUpClient() as client:
    meetings = client.meetings.list_meetings(
        ListMeetingsRequest(status="analyzed", limit=25, offset=0)
    )

    analyses = client.meetings.list_meeting_analyses(
        ListMeetingAnalysesRequest(
            meeting_ids=["MEETING_ID_1", "MEETING_ID_2"],
            include_deal_intelligence=True,
            include_feedback=True,
        )
    )

The client reads AMPUP_API_KEY and AMPUP_BASE_URL from the environment. See the Python SDK guide for the full client surface.

What a GTM agent can retrieve

The exact surface depends on user permissions and enabled products.

Context	Start with	Go deeper when needed
Workspace	`get_workspace_overview`, `list_integrations`	`list_organizations`, `select_organization`
Accounts and people	`list_accounts`, `list_contacts`	`get_account`, `get_contact`
Pipeline	`list_deals`	`get_deal`, deal tasks and notes
Meetings	`list_meetings`, `get_meeting`	transcript excerpts, full transcript, briefs
Analytics	`list_evaluation_metrics`	bulk meeting analyses, deal intelligence
Coaching	`list_roleplays`, `list_courses`	roleplay sessions and assignments
Prospecting	`get_icp`, `list_audiences`, `list_campaigns`	sequences, analytics, conversations

MCP tool names evolve as related operations are consolidated. For an MCP connection, use tools/list for the live schema. For direct API and SDK integrations, pin a tested client version and validate response shapes before promoting a pipeline.

Five starter journeys

1. Keep the CRM current

List newly analyzed meetings.
Get the meeting overview and supporting evidence.
Resolve the account, contacts, and deal.
Draft notes, tasks, and field updates.
Require human review before CRM sync or a stage change.

2. Build an ICP list

Use the MCP workflow tools to search companies and people, enrich missing details, score ICP fit, save the qualified records to an audience, and draft a sequence. Keep launch disabled until a person reviews targeting and copy.

3. Feed conversation data to BI

Use the direct API to page through analyzed meetings, bulk-fetch analyses in batches of at most 100 IDs, normalize metric and deal-intelligence fields, and checkpoint the last successful window before loading the warehouse.

4. Close the coaching loop

Find low-scoring meeting moments, create a meeting-derived roleplay, propose an assignment, and read the resulting sessions back into the coaching workflow.

5. Act on external signals

Reconcile the incoming company and contact, attach the signal as prospecting context, and draft signal-aware outreach. Require approval before sending.

Production pattern for data pulls

Discover narrowly. Filter list calls and page with limit and offset.
Hydrate selectively. Fetch details only for new or changed IDs.
Batch analysis. Use list_meeting_analyses instead of one request per meeting.
Avoid unnecessary transcripts. Start with get_meeting; pull excerpts or the full transcript only when exact wording is required.
Checkpoint progress. Persist the last successful window or record IDs so retries do not restart the entire job.
Retry safely. Back off on 429 and transient 5xx responses; surface persistent validation and authorization failures to an operator.

Errors and troubleshooting

Status	Meaning	Action
`401`	Missing, invalid, or revoked credential	Verify the bearer header or reconnect OAuth
`403`	The user or role cannot perform the operation	Confirm workspace role and tool permissions
`422`	Request body does not match the current schema	Re-read the MCP tool schema or SDK request model
`429`	Too many requests	Back off with jitter and retry
`5xx`	Temporary service failure	Retry a bounded number of times, then alert

Next steps

MCP server reference

Endpoint, client setup, authentication, and the current tool catalog.

Security model

OAuth flow, token lifetimes, permissions, tenant isolation, and revocation.

Python SDK

Typed request models and a high-level client for server-side jobs.

TypeScript SDK

Generated API clients for TypeScript services.

Need help? Email support@ampup.ai.

Start a free workspace

Open the MCP reference

​1. Create and authorize the workspace

​2. Connect an interactive agent with MCP

​3. Call the API directly

​List analyzed meetings

​Get one meeting and its analysis

​Bulk-pull analysis for a warehouse job

​Pull the full transcript only when needed

​Pull deal-level conversation intelligence

​Server-side JavaScript helper

​Python SDK equivalent

​What a GTM agent can retrieve

​Five starter journeys

​1. Keep the CRM current

​2. Build an ICP list

​3. Feed conversation data to BI

​4. Close the coaching loop

​5. Act on external signals

​Production pattern for data pulls

​Errors and troubleshooting

​Next steps

MCP server reference

Security model

Python SDK

TypeScript SDK

1. Create and authorize the workspace

2. Connect an interactive agent with MCP

3. Call the API directly

List analyzed meetings

Get one meeting and its analysis

Bulk-pull analysis for a warehouse job

Pull the full transcript only when needed

Pull deal-level conversation intelligence

Server-side JavaScript helper

Python SDK equivalent

What a GTM agent can retrieve

Five starter journeys

1. Keep the CRM current

2. Build an ICP list

3. Feed conversation data to BI

4. Close the coaching loop

5. Act on external signals

Production pattern for data pulls

Errors and troubleshooting

Next steps