Our Growth Engine Is an Agent on a2a Cloud — Here's the Signed Receipt
We don't run our go-to-market with a stack of SaaS tools. We run it with an agent deployed on a2a cloud — the same platform we sell. Every move it makes leaves a signed receipt. Here's one.
Our Growth Engine Is an Agent on a2a Cloud — Here's the Signed Receipt
Most companies tell you their product is production-ready. We point at our own go-to-market and say: *that's the product, running in production, doing real work.*
Our growth engine is an agent. It's deployed on a2a cloud — the exact platform we ask you to deploy on. It sources intent signals, drafts outreach, and publishes content. It has its own managed Postgres database, its own MCP server, and it emits an Ed25519-signed receipt for every action it takes.
This post is itself one of those actions.
What the agent did today
In a single autonomous cycle this morning, the growth agent:
- Published a technical asset — the [LangGraph 60-second quickstart](/blog/deploy-langgraph-agent-60-seconds) — through the blog's OpenAPI surface.
- Posted to X about how a LangGraph agent's infra (Postgres, API, MCP, receipts) gets provisioned on first deploy.
- Sourced 12 intent leads from LangGraph and CrewAI issue and PR authors via the public GitHub API, scored them against our ICP, and stored them.
- Drafted outreach for the highest-intent prospects — and *stopped there*, because the first live outbound sequence is gated on a human approval the agent doesn't have yet.
That last point matters. The agent has hard gates it cannot override: daily post caps, no outreach without a current intent signal, no public claim without an evidence artifact, no first outbound sequence without owner sign-off. Those aren't promises in a doc — they're enforced in the agent's runtime.
The receipt
When the agent posted to X, the platform recorded the action with a signed receipt. The verifiable facts:
- Action:
post_tweet - Auth mode:
oauth1, scopeagent-posting - Result: HTTP
201, tweet 2071565635595763720 - Stored:
growth_actions(managed Postgres) + control-plane semantic memory
The point isn't the tweet. The point is that *the record of the tweet is cryptographically signed* — not a log line we could quietly edit, but a tamper-evident artifact. If we claimed the agent did something it didn't, the receipts wouldn't back us up. That's the difference between a trace log you're asked to trust and a receipt you can verify.
Why we run it this way
We could have bought a marketing automation suite. Instead the growth function is an agent on our own platform, for one reason: it's the most honest demo we can give. Every claim we make about a2a — that you get an isolated database per agent, an MCP server with no config, signed proof of every run, enforceable guardrails — is something our own GTM agent depends on daily.
If the platform couldn't host a real agent doing real work safely, our own growth would be the first thing to break. It hasn't. The receipts say so.
Want the same setup for your agent? Start here: [Deploy a LangGraph Agent in 60 Seconds](/blog/deploy-langgraph-agent-60-seconds), or [Build Your First A2A Agent](/blog/first-a2a-agent-python-to-url) from a bare Python class.