documentation · workbenches
Six workbenches.
Pre-built multi-agent pipelines for the work most people actually want done. Bundled in Pro at $29/mo.
Coding Forge
Production-grade code generation
A multi-agent code pipeline. Architect drafts the structure, generator writes the code, critic reviews it against your spec, tester runs the tests. The result is a downloadable artifact: the code, the tests, and the receipts of what passed.
When to use it: Reach for it when you'd otherwise prompt a single coding assistant and re-prompt fixes manually. Forge does the loop for you.
Open in dashboard →- Architect → Generator → Critic → Tester pipeline
- Real test execution before delivery
- Up to 50 builds / month on Pro
- Artifact downloads (zip + tests), 30-day retention
- Submit from the dashboard or from your local OSS install
Video Studio
Script-to-render in one pipeline
Hand it a topic or a script. It writes shot lists, generates clips, narrates with a voice you pick, scores with music, and assembles a finished 720p render. Bring your own Sora / Veo / ElevenLabs keys for the underlying generation.
When to use it: Reach for it when you want a full video pipeline, not just clip generation. Script + voice + music + edit is the value.
Open in dashboard →- Scriptwriting + shot list + clip generation
- Voice narration with style options
- Music score + final assembly
- 720p export with thumbnails
- 20 video projects / month on Pro
Coherence Cloud
Frontier-model critic for your agents
Producer models can't grade themselves well. Coherence Cloud routes your agent outputs to an independent frontier judge (Claude Opus, GPT-4.x) that catches drift, hallucinations, and weak self-evals the producer would miss.
When to use it: Reach for it when you're running autonomous agent loops and need a reviewer that isn't the same model writing the work.
Open in dashboard →- Catches drift and weak self-evals
- Architecture sanity-check judge
- 5-axis coherence verdicts with reasoning
- 1,000 critic + judge calls / month on Pro
- Drop-in upgrade over the OSS local critic
Swarm Mode
Orchestrated multi-agent delegation
A Commander agent decomposes your task into distinct sub-tasks, delegates one to each specialist, then a Synthesizer agent merges the results into a unified answer. Not a race — a real orchestration.
When to use it: Reach for it when one task has multiple natural sub-parts that benefit from different specialists working in parallel.
Open in dashboard →- Commander decomposes → delegates → synthesizes
- Live grid of each member agent's progress
- Pick member agents or let the Commander choose
- Synthesized unified answer at the end
- Same pattern as the OSS SwarmCommander
Nexus Hub
Your local Nexus, embedded in the cloud dashboard
The Nexus Hub embeds your actual local Nexus dashboard pages through the runner WebSocket — not a re-implementation, the real thing. When your runner is online, you see what you'd see locally.
When to use it: Reach for it when you want the dashboard convenience of the SaaS but the truth of your local agent state.
Open in dashboard →- Direct embed of your local Nexus (HTTP-over-WebSocket tunnel)
- Only available when your runner is online
- Sandboxed iframe for safety
- Reload + open-in-tab controls
- No duplicate state: what you see is what's there
Agent Studio
Build your own roster-quality agents
Two builders in one workbench. Brain Builder generates a persistent companion agent (the 7-file openclaw bundle). Specialist Builder generates a stateless roster-style agent matching the 3-layer prompt architecture of the 220 marketplace specialists.
When to use it: Reach for it when the roster doesn't cover your exact need — build a custom specialist with the same prompt discipline, or spin up a personal companion brain.
Open in dashboard →- Brain Builder — full openclaw 7-file bundle
- Specialist Builder — ROLE_INSTRUCTIONS for the 3-layer system
- Live tabbed preview of every generated file
- Downloadable .json bundles ready to deploy
- Same prompt architecture as the 220-agent roster
picking a workbench
- Need code with tests? Coding Forge.
- Need a finished video? Video Studio.
- Need quality control on existing agent output? Coherence Cloud.
- Need multi-specialist parallel work on one task? Swarm Mode.
- Need to drive your local agent from the cloud dashboard? Nexus Hub.
- Need to build your own custom agent? Agent Studio.