fl0w · AI Workflows
Tune your workflows locally. Run them anywhere.
fl0w is an AI Workflow product — not an “AI coworker.” Every tool, MCP server, and skill load becomes a step in a workflow you build in your CLI with jaaw. Then run your workflows on a schedule or via API, wherever you need them.
workflow · daily-digest
09:00 dailylistAllTools()
Gather the tools available
mcp · github.search
Jev picks the next step
skill · summarize
Load a skill as a step
stop
Jev decides it's done
One workflow, two places
Tune it on your machine. Ship it to ours.
The workflow you perfect in your terminal is the same one that runs in production. Nothing gets rebuilt in between.
Locally · jaaw
Build and tune in your CLI.
jaaw runs a loop that calls Jev until it chooses to stop. Every other choice is a tool you can inspect and adjust.
- Your MCP servers, skills, and built-ins as steps
- Watch every choice Jev makes
- Open source — install in one command
Remotely · fl0w
Run it wherever you need it.
Push the workflow you tuned to fl0w. It runs on our infra with the exact steps you perfected — no rework.
- CRON schedules
- Trigger via API
- Same MCP tools and skills as local
The 0 in fl0w
Building prod, not God.
When you read human-like responses from LLMs, it's easy to think you're looking at some higher level of 'intelligence.' You're not. You're looking at a machine that interprets your intention and executes the right steps — your routine, on fire.
01
Not an AI coworker
LLMs make choices based on statistics — and with the right sequence of choices and tools, they can complete the tasks you give them. That's a workflow, not a colleague.
02
AI Workflow, not AI intelligence
Human-like responses fool you into thinking there's something 'higher level' going on. There isn't. It's a sequence of choices and tools executing your routine.
03
You stay in control
You tune every step locally. Each tool, MCP tool, and skill load is a node you can inspect, adjust, and perfect before it ever runs remotely.
“In many ways it’s just like a traditional agent harness — but putting Jev in the driver’s seat of a task loop feels less like an intelligence and more like a machine that can interpret your intention and execute the right steps to complete it.”