GTM infrastructurefor coding agents
Your coding agent writes your GTM workflows as code. Cargo runs them, like Vercel runs a web app.
Trusted by GTM teams at
Most GTM stacks are ten toolsheld together by one person
AI made GTM workflows easy to build. Keeping them working is the hard part: they depend on one person, they stall when a quota or an API changes, and when a lead goes to the wrong rep nobody can say why.
Your stack today
- A different tool for each step
- Only one person knows how it works
- It breaks when an API or a quota changes
- Nobody can say why a lead went where it did
On Cargo
- One system does every step
- The logic is code anyone can read
- Cargo runs it and retries what fails
- Every run is logged, step by step
Northwind
Owner updated
Message sentWhere Cargo sits in your stack
For GTM engineers, RevOps and revenue teams. Cargo reads from your tools and writes the results back where your team works.
What Cargo reads
The data you already have, wherever it lives today.
- HubSpot: Your CRM
- Salesforce: Your CRM
- Snowflake: Your warehouse
- BigQuery: Your warehouse
- Postgres: Your warehouse or product database
- Stripe: Billing
- Amplitude: Product usage
- Mixpanel: Product usage
- Snitcher: Website visitors
- Forms: Website forms, by webhook
- Calendar: Meetings, from Google Calendar
- Signals: Cargo Signals: job changes, funding and hiring
- Apollo: Company and contact data
- Sheets: Google Sheets and CSV files
- 100s more
What runs in Cargo, as code
Models, context, workflows (plays), agents and tools.
Your repo- my-gtm/
- context/
- icp/mid-market.mdICP
- persona/revops.mdpersona
- infra/
- models/companies.tsdata
- plays/inbound.tsworkflow
- tools/enrich.tstool
- agents/research.tsagent
Where results go
Back into the tools your team already works in.
- HubSpot: Fields, scores and owners, back in your CRM
- Salesforce: Fields, scores and owners, back in your CRM
- Slack: Alerts and approvals for your team
- Email: Sent from your reps' own mailboxes
- Outreach: Your sequencer
- lemlist: Your sequencer
- Instantly: Your sequencer
- Claude: Your tools and agents, over MCP
- ChatGPT: Your tools and agents, over MCP
- Chrome: The Chrome extension, next to the page your rep is on
- Apps: Internal apps for your reps, hosted on Cargo
- Button: Reps run a tool from the record
- Sheets: Rows written to Google Sheets
- Webhooks: Results sent to any other system
- API: Read results from your own code
Your tools stay. What you stop building and maintaining is everything between them: the glue, scheduling, retries and rate limits, and tracing what went wrong.
What it looks like in the app
Data models
Your CRM, warehouse, product data and enrichment merged into one set of companies, contacts and deals.
Context
Context is what your team knows, written down as files: your ICP, personas, objections and rules. Every agent reads the same files.
Workflows
Cargo calls these plays. A play runs on a trigger, like a new HubSpot contact, or on a schedule. Cargo retries failed steps and stays within API limits.
Mailboxes
Cargo sends from your reps' Google and Outlook mailboxes, warms each one up and keeps it under its daily limit.
Run logs
A run is one record going through a play. Every run is logged step by step, so you can see why a lead went where it did.
Projects
Your plays, agents and context as TypeScript files in a git repository. Your coding agent edits them, a person reviews the pull request, and merging deploys it.
How it works
GTM as code: your workflows are TypeScript files in a git repo, changed by pull request. Merging deploys them.
1. Agent writes it
Claude Code, Cursor or Codex edits your files.
infra/plays/inbound.tsexport const inbound = definePlay("inbound", { model: contacts, workflow: routeLead, });2. You review it
The pull request lists what will change.
$cargo-ai project plan+ play inbound+ workflow route-lead~ model contacts2 to create, 1 to updateposted on PR #423. Merge deploys it
Cargo runs the new version and logs every run.
$cargo-ai project deploydeployed, 3 resourcesrun #1 Northwind scored 82 / 100 routed to Alicev12 kept to roll back4. Agents improve it
Agents read the run logs and open a pull request.
Raise min scoreopened by an agentafter 1,240 runs- minScore: 70+ minScore: 75awaiting your approval
Nothing goes live until you approve it. Each approved edit goes through steps 2 and 3 again.
What teams use Cargo for
Each of these is a cookbook: a working example your coding agent adds to your repo and adapts to your team.
Teams that run their go-to-market on Cargo
“Most ICP models are built once and forgotten. Ours runs perpetually: every closed-won, every churn, every pattern. Cargo enriches the gaps, does the analysis, and deploys it as a skill we can call from anywhere. This is what infrastructure actually means.”
Alexis Girard · Pivot
Build it yourself, or have our team build it
Both options run on the same Cargo account and the same repo. If our team builds it first, your team can take it over later without moving anything.
- For GTM engineers and their coding agents
Start building
Your coding agent writes the workflows, and Cargo runs them in production.
- Your GTM as code, in a repository you own
- Data models, context and workflows, ready for your coding agent
- Free credits to start, no card
- For revenue leaders and teams that want it built and run for them
Build with us
Our team builds, operates and continuously improves it with you.
- Working GTM workflows, delivered as a repository you own
- A dedicated Slack channel, onboarding and on-demand calls
- Your engineers can take it over at any time
Enterprise-grade infrastructure
SOC 2 Type II certified
Audited every year by independent third parties.
Audit log & evals
Every run and approval is logged. Evals gate prompt changes.
Data residency
Choose the region where your data lives.
SSO & role-based access
SAML single sign-on and role-based permissions.
Version history
Every play and agent is versioned, with diffs and rollback.
GDPR & trust center
DPA, sub-processor list and security docs for procurement.
Frequently asked questions
What Cargo is, how you build on it, and how to start.