🚀 All Automations 109
🦞 Powered by Lobster
The open-source workflow engine that makes ClawFlows automations actually run.
Without Lobster
LLM orchestrates every step
10+ tool calls, ~100K tokens per run
With Lobster
lobster run workflow.yaml
Deterministic pipeline, 0 LLM tokens
Zero Token Cost
Data collection runs as shell commands — curl, jq, scripts. No LLM needed to fetch calendars, check APIs, or parse RSS.
Stateful Pipelines
Workflows remember what they've processed. Yesterday's changelog check won't re-alert on old releases.
Approval Gates
Workflows halt at checkpoints and ask before acting. No "oops, it sent 50 emails" moments.
Lobster is the execution engine. ClawFlows is the registry.
Together: npm + node for AI agent automations.
⚡ Two Formats, One Registry
Every automation comes in two flavors: abstract (portable) and Lobster-ready (runs now).
📋 Abstract (Capability-Based)
needs: prediction-markets
Portable — works with any skill
🦞 Lobster-Ready (Concrete)
command: curl ... | jq ...
Runs NOW — zero LLM tokens
Abstract format = universal spec for any agent runtime. Lobster format = deterministic shell pipelines that execute today. Use abstract for portability, Lobster for performance.
Standard Capabilities
These are the building blocks. Browse all or propose new ones.
prediction-markets
Polymarket, Kalshi
database
SQLite, storage
chart-generation
Graphs & visuals
social-search
X/Twitter search
calendar
Events & scheduling
tts
Text to speech
🤝 Add Your Automation
Just tell your agent what you want automated. It handles everything — writing the YAML, creating the PR, the whole thing.
"Create a ClawFlows automation that tracks [X] and alerts me when [Y]. Publish it to the registry."
Your agent will:
- 1 Write the automation YAML with the right capabilities
- 2 Create README with examples and documentation
-
3
Open a PR to
clawflows-registry
Manual instructions in the Publishing Guide if needed.
