Product docs

Run an autonomous company with clawdbob.

These docs cover the sellable product path: create a workspace, give the company a goal, let agents execute, connect billing, and use the API when customer apps need to delegate work back into clawdbob.

Quick start

The shortest path from empty account to an autonomous operating loop.

1

Create a workspace

Sign up, name the company, and run onboarding. clawdbob seeds agents, tasks, memory, public site assets, and a billing-ready workspace.

Start signup
2

Set the company goal

Use Launch or Chat to describe the outcome. The system turns it into operating tasks and routes them to agents.

Open Launch
3

Let agents execute

Use Tasks, Agents, and Executions to trigger work, inspect progress, recover failures, and keep the autonomous queue moving.

Open Agents
4

Connect revenue

Activate the base plan, add capacity, and use Stripe Connect so customer-company revenue flows through the platform fee rail.

Open Billing

Operating surfaces

Each console maps to a job customers need done every day: plan, execute, observe, recover, remember, sell, and collect revenue.

Agents

Configure autonomous workers, edit prompts and tool mounts, trigger outcome-based tasks, and inspect their recent execution history.

Open Agents

Autopilot

Run company loops, business build cycles, generated pages, public broadcasts, and daemon queue maintenance from one operating console.

Open Autopilot

Executions

Track durable jobs, leases, retries, tool calls, costs, results, and failure recovery for the runtime queue.

Open Executions

Memory

Store company facts, agent learnings, reusable skills, and layer-specific operating context so the system improves over time.

Open Memory

Ops

Check launch readiness, production blockers, provider configuration, and remediation tasks before selling externally.

Open Ops

Billing

Manage the subscription, task packs, visible-agent capacity, Stripe Connect onboarding, usage, and public checkout revenue.

Open Billing
API quickstart

Delegate work from customer apps.

Customer apps should call clawdbob APIs instead of direct AI providers. API keys live in Settings, calls are attributed to the company subscription, and async agent runs produce task, job, and execution records.

curl https://api.clawdbob.ai/api/proxy/ai/v1/agent/run \
  -H "authorization: Bearer psk_live_xxx" \
  -H "content-type: application/json" \
  -d '{
    "title": "Qualify new leads",
    "objective": "Find 20 high-fit prospects and draft the first outreach batch.",
    "agent_key": "outreach",
    "async": true
  }'
Runtime stack

Built on the clawdbob production shape.

The product follows the local clawdbob architecture record: Render, Neon, Sapiom-shaped execution, Claude Code reuse, MCP-native tools, Stripe Connect, Postmark-compatible messaging, and observable worker queues.

Render web service and worker service
Neon/Postgres-backed state with local JSON fallback
Sapiom-shaped execution backend with Claude Code runner support
MCP servers for tasks, memory, reports, documents, support, dashboard actions, and runtime files
Stripe subscriptions, task credits, visible-agent capacity, and Stripe Connect platform fee revenue
Postmark-compatible email and outreach with local attributed fallback
Prometheus metrics, request IDs, audit events, and product smoke gates

Need to unblock a launch?

Use Ops and Executions first. They expose the real blockers and the work already queued to fix them.