Operators of AI-native businesses
Work that needs to touch the physical world and collaborate across agents / tools / machines — not just a coding sandbox.
Long-running Workflow Harness
Make your Workflow Clawsome!
ClawsomeFlow isn't here to replace the agents or workflows you already have — it only makes them more robust and extensible!
One Platform. Infinite Possibilities.
Who should try it?
Built for work that leaves the AI sandbox
Builders done with mega-context chats
Want flows that are reusable, auditable, partially re-runnable, and resumable from checkpoints — cost-controlled and dependable over long cycles.
Core features
Core product features
Fuse many kinds of executor
Break out of local-only agent collaboration. Real people, CLI agents, remote ClawsomeFlow instances, custom agent tools — anything you could call an execution unit — share one graph, with the same dependency and completion semantics.
Human + machine · cross-agent platforms
People, remote ClawsomeFlow, webhooks, and CLI agents sit in the same DAG.
Hand off across machines
Delegate a subtask to a Flow on another machine.
Bring your own executor
Wire in any system you built through a tiny webhook contract — anything that can report a result can be a node.
Harness engineering
Purpose-built for ultra-long runs: pause anytime, resume anytime, review and re-run locally, let an agent pause itself, and fix a single failure without restarting the whole flow — so long task chains stay on target.
Cost-aware by design
Split work across nodes so each executor sees a short, relevant context — far cheaper than one mega-agent hauling the whole project.
A process you can steer
Human checkpoints, partial subtask re-runs, agent self-pause, auto-pause on failure, and a complaint-driven loop — coordination lives in the harness, not a hopeful prompt.
Repo lock & isolation
Worktree isolation plus a built-in cross-process repo lock keeps parallel edits off each other's toes; merges and results stay reviewable and revertible.
Observable end-to-end
Every dispatch, hand-off, and failure is a RunEvent — a live board, inbox messages on dependency edges, and full replay. No black boxes.
Reusable, repeatable flows
Define a Flow once and re-run it with parameters — same structure, predictable delivery, not a throwaway prompt session.
Built for long cycles
Wait on people or remote jobs for hours or months without losing the thread — pause, resume, complain, rework, and keep getting better.
How it works
From a sentence to a shipped result
You stay in charge of the goal. The harness coordinates every execution unit — and recovers when something stalls, without throwing away the whole run.
-
1
Describe your goal
Tell ClawsomeFlow what you want in plain language, or compose a Flow visually as a graph of tasks and dependencies.
-
2
Agents run in parallel
The scheduler actively dispatches ready tasks to the right agents, each in its own isolated workspace, and drives them to completion.
-
3
Watch, steer, recover
Live board, checkpoints, selective re-run, pause/resume, retry/skip/abort — fix one point without restarting all.
-
4
Converge & deliver
A leader merges the parallel work into one reviewed deliverable — and the run history stays fully auditable.
Works with your stack
Bring the agents you already love
We don't replace your agents or workflows — we make them more robust and extensible. CLI agents, humans, remotes, and tools share the same graph.
OpenClaw
Hermes
Claude Code
Codex
Cursor
Gemini CLI
Kimi CLI
Qwen Code
OpenCode
Qoder CLI
CodeBuddy Code
nanobot
Quick start
Up and running in one line
Install the CLI (Linux / macOS)
curl -fsSL https://clawsomeflow.com/install.sh | bash
Join the community
Let's build clawsome things together
Questions about using ClawsomeFlow, or curious about building a one-person company powered by agents? Join our Discord server or scan the QR code to join our WeChat discussion group.
