Agent Flow
A visual canvas where a node is a live coding agent and an edge is a two-way message bus. Add automation triggers and a receipt for every run, and you get the one category nobody owns yet: visual, live, communicating, trigger-driven, and accountable — in one place.
Every competitor owns one axis.
Nobody owns the intersection.
A visual canvas where a node is a live coding agent and an edge is a two-way message bus — with automation triggers and a receipt for every run. Design it and supervise it on the same graph.
| Product | Visual canvas | Live coding agents | Two-way agent comms | Automation triggers | Receipts / monitor |
|---|---|---|---|---|---|
| Zapier / Make / n8n | |||||
| cmux | |||||
| LangGraph Studio | |||||
| CrewAI / AutoGen(code-only) | |||||
| Bootspring Agent FlowEarly access |
Competitor capabilities as of July 2026, based on public docs. “Receipts / monitor” means per-agent run receipts and reproduce-vs-cut signals, not generic execution logs. Bootspring Agent Flow is in active development — see what ships when.
How it ships
The engine comes before the canvas. We build the fleet you can drive, then the surface you draw it on — and we label what is live versus what is next.
Fleet backend ("our own tmux")
Drive real agents in parallel — tmux control mode or an owned PTY manager, cross-platform.
Control loop
Programmatic access to every agent: spawn, send, read, key, close, broadcast, shard, race, watch.
Two-way comms harness
Any agent can message any agent or the orchestrator — a durable mailbox bus, not a black box.
Fleet presets
Three-tier, race, shard, review-panel, swarm-dispatch — the fan-out/fan-in patterns, one command.
Monitor-to-improve
Score every run, surface which agents to reproduce and which are burning tokens.
Agent Flow canvas
Our own canvas engine — nodes are live agents, edges are the message bus. Design your fleet visually.
Live overlay
The same graph becomes the cockpit — nodes ring, stream output, show cost, in real time.
Trigger-driven fleets
PR opened → spawn a review fleet → race a fix → post to Slack → human approval → merge.
Replay + receipts
Time-travel any run on the graph. Every node carries a receipt: who ran, what it cost, who won.
Deep-dive comparisons
Honest, side-by-side looks at each axis of the category.
Frequently asked
- Is Agent Flow available today?
- No. Agent Flow is in active development and offered as early access. Wave 1 (the agent fleet engine) comes first; the visual canvas (Wave 2) builds on it. We label what ships when rather than overselling.
- How is this different from Zapier, Make, or n8n?
- Those automate API calls between apps. Agent Flow nodes are live coding agents that talk to each other and can be launched by the same kinds of triggers — so an automation can spin up a fleet of agents, not just move data.
- How is this different from cmux?
- cmux is a great Mac-only terminal for driving parallel agents, but there is no visual canvas, no automation triggers, and no run receipts. Agent Flow is cross-platform, adds the canvas and triggers, and records a receipt for every run.
- How is this different from LangGraph Studio or CrewAI/AutoGen?
- LangGraph Studio visualizes an LLM graph but not live terminal coding agents; CrewAI/AutoGen are code-first with no visual canvas or automation surface. Agent Flow combines the visual canvas, live agents, two-way comms, triggers, and receipts in one place.
- What does "receipts / monitor" mean?
- A per-run record — which agents ran, what each was asked, tokens and cost per agent, who won a race and why, and gate/QA outcomes — plus reproduce-vs-cut signals. Not generic execution logs.
Put a control plane in front of every agent.
Create an account, install the CLI, and run the build loop against a real project. todo.md stays honest. Receipts stay visible.
npm i -g @girardmedia/bootspring