Overview
Clockwork is a workflow execution platform. You define task graphs — who does what, in what order, with what constraints — and Clockwork continuously schedules, re-schedules, and monitors them as the real world changes.
The core model
Every piece of work in Clockwork is one of these:
| Concept | Description |
|---|---|
| Task | A single unit of work with a duration, assignee, and deadline |
| Workflow Instance | A running collection of tasks forming a directed graph |
| Workflow Template | A reusable blueprint that stamps out instances |
| Dependency | A link between two tasks enforcing execution order |
| Resource | A person, agent, service, or asset that executes tasks |
When a task completes, Clockwork automatically re-schedules every downstream task, re-computes the critical path, and surfaces any new conflicts or scheduling opportunities.
What you can build
- Automated pipelines — Create and manage workflow instances programmatically as your system processes events
- Human + AI workflows — Mix human approval gates with AI agent tasks in a single graph
- Monitoring dashboards — Stream real-time task events to your UI or alerting system
- AI agent integration — Connect Claude, ChatGPT, or any MCP-compatible agent to Clockwork's full toolset
Which integration to use
| Interface | Best for |
|---|---|
SDK (@clockwork/sdk) | Server-side automation, custom integrations, scripting |
| REST API | Any language, simple webhooks, raw HTTP |
MCP (@clockwork/mcp) | AI agents using the Model Context Protocol (Claude, etc.) |
The scheduling engine
Clockwork runs a Critical Path Method (CPM) engine on every mutation. You don't need to think about scheduling — just define tasks and dependencies, and Clockwork tells you:
- Which tasks are on the critical path (
isCritical: true) - How much float each task has (
slackSec) - Which constraints are violated (
conflicts) - Which optimization windows are available (
opportunities)