Repo-aware launch pack
Reads the diff, README, routes, pricing, and product context before writing the launch.
online
Business execution layer · GTM first
Your agent built it.
ChiefLab launches it.
ChiefLab is an MCP execution layer your coding agent calls when you say launch this. It reads the repo, picks channels for what you built, creates a review URL, stages publish/send actions, refuses to execute before approval, then measures results 24 hours later.
One install. Your agent gets the API key automatically on first call. Approval required before any publish, send, or paid execution.
Live endpoint
https://api.chieflab.io/api/mcp1 commandfrom build to launch room
0posts without approval
24hresults readback
Hosted MCP live · approval enforced server-side · npx @chieflab/cli live on npm
> user: "launch this"
reading: diff, README, routes, pricing
created: launch pack + chieflab.io/review/run_8f31
waiting: human approval before publish/send
24h later: results readback + next move
Repo launch pack Diff-grounded
Approval room Review URL ready
Publish actions Staged, not fired
Next move Measured after launch
approval room chieflab.io/review/run_8f31
awaiting human 24h Next move queued
Live activity Verify any of these →
2026-05-04 Dogfood — chieflab.io launch Zernio post 69f8a74f… · Resend ba122a9c… approved + executed
2026-05-07 Customer smoke (production) 105/105 checks · api.chieflab.io passing
2026-05-10 Hero + audit funnel deploy chieflab.io build · 44 pages live
The first business loop
Install once. Your agent gets a GTM operating layer.
Coding agents can ship products faster than founders can run the business around them. ChiefLab starts with the first loop every agent-built company needs: launch, approval, execution, measurement, and memory from Cursor, Claude, Codex, Lovable, Bolt, Replit, CLI, or direct MCP.
Approval room
Creates a signed review URL your human can open, edit, approve, or reject.
onlineSocial actions
Stages LinkedIn, X, Product Hunt, and follow-up posts without firing them early.
onlineLaunch email
Drafts the email sequence and sends through Resend only after approval.
onlineCreative briefs
Turns the product context into image prompts, ad angles, and reusable campaign assets.
online24h measurement
Reads launch results and recommends the next move instead of stopping at copy.
online Control layer, not autopilot
Agents prepare the business action. Humans approve.
ChiefLab is not autopilot. Agents prepare external work, ChiefLab checks readiness, and humans approve before anything posts, sends, or spends.
Drafts from repo contextLaunch copy starts from the product, not a blank template.
Stages every external actionPublish and send tools refuse until the action is approved.
Measures the next moveResults feed the follow-up recommendation and workspace memory.
CL
chieflab.io/review/run_8f31 Awaiting human approval
pending LinkedIn founder storyReady
X launch threadReady
Launch emailReady
24h measurement planQueued
GTM first
Start with launch, not a fake all-in-one promise.
ChiefLab earns trust by doing one urgent job well: turning a shipped product into a reviewed launch. Workspace memory, connectors, approval history, and client keys come after the first launch loop works.
1 command from shipped product to launch room
0 posts or emails without approval
Zernio + Resend real execution after review
Founder @ AgentBuiltCoLinkedIn draft
We shipped with an agent, then asked ChiefLab to launch it. It read the repo, wrote the founder story, staged the publish actions, and waited for approval.
87 reactions14 comments12 signups
Your product agent can launch now
The product is built. ChiefLab is preparing the launch pack, approval room, publish actions, and 24-hour readback.
Choose install path → How it works
From shipped product to launched product.
One agent instruction starts the loop. ChiefLab handles the launch work around the model: approvals, assets, connectors, measurement, and memory.
-
01 Your agent sends the product context
Recent changes, routes, README, screenshots, pricing, and the outcome you want. ChiefLab starts with what actually shipped, not a generic URL scrape.
-
02 ChiefLab creates the launch room
Positioning, channel assets, creative prompts or images, connector readiness, publish/send actions, and a signed approval link for the human.
-
03 You approve what goes out
No hidden publishing. Approve, reject, or edit each action. If a channel is not connected, ChiefLab shows the blocker and gives a manual fallback.
-
04 ChiefLab measures and improves
After launch, ChiefLab reads engagement, traffic, queries, and sends. The next launch reuses the approved copy, objections, channel results, and winning angles.
Want the tool contract? Read the full agent instructions →
The platform primitive
Agents can move fast.
ChiefLab makes external business work safe.
Guardrail What it prevents
Human approval
Every external action waits for explicit approve, reject, or edit.
Connector readiness
Missing Zernio, Resend, GA4, Search Console, or CMS setup becomes a clear next step, not a fake success.
Audit trail
Runs, assets, actions, approvals, and rejection feedback stay attached to the workspace.
No surprise cost
Hosted image generation, publish, send, and full server-side generation are gated before execution.
That is the product boundary: agents propose and prepare; humans approve; ChiefLab executes only what is approved. Today that means GTM. The same trust model can support more business functions later.
Install the post-build launch loop.
Pick your runtime first. Sign in only when you need workspace memory, approval history, connectors, or a key for a client that requires one.
install in: Cursor · Claude Desktop · Codex · HTTPS · all runtimes
Need a key for a client config? Sign in and manage it from the workspace.
FAQ
The questions worth answering up front.
Is ChiefLab a tool or a service?
The interface is a tool your agent can call. The product is a launch execution loop: ChiefLab turns repo context into launch strategy, assets, approval, publish/send execution, measurement, and the next move. GTM is first because every shipped product needs users.
Do I have to use Cursor or Claude Desktop?
No. ChiefLab is a hosted JSON-RPC endpoint at chieflab.io/api/mcp. Anything that can POST HTTPS with a Bearer key calls it: a Telegram bot, a Vapi voice agent, a Lovable web app, a LangChain script, your own SDK. Cursor + Claude Desktop are first-class via native MCP, but they're two of many.
Does ChiefLab write the final assets or just briefs?
Both modes exist. Default context mode returns repo-grounded briefs your agent's LLM renders. Repo-aware launches can auto-promote to draft mode for reviewable copy. If you want ChiefLab to render server-side, use full mode. Either way, the approval room shows what the human is approving before anything external fires.
Why won't I just ask Cursor or Claude to launch this?
The model can draft copy. It usually cannot safely own repo-grounded memory, connector auth, approval states, signed review URLs, idempotency, publishing, send logs, 24h readback, and repeat-launch learning. ChiefLab wraps those operational pieces into one service loop your agent can call.
Can I self-host?
Not in v1. ChiefLab runs as a hosted JSON-RPC endpoint at chieflab.io/api/mcp. The hosted runtime, OAuth token vault, brand-context cache, approval state, and connector layer are what you're paying for — re-implementing them defeats the point. If you have a hard self-host requirement, email hi@chieflab.io.