Phorma Labs

Build agents you can watch think.

Phorma Agent Studio is the visual canvas for multi-agent AI. Compose agents as nodes, wire their handoffs, then run the workflow and watch it execute live — nodes glow while thinking, edges pulse on handoffs, tokens stream in real time. Pause, step, and set breakpoints like an IDE. Deploy the exact same workflow to your own cloud, unchanged.

Free for individuals · No credit card required

support-triage.workflow

Running

Router

Classifies incoming requests and routes to the right agent.

gpt-5 · 41ms

Researcher

Pulls context from your docs, tickets, and knowledge base.

Web Search

Runs live web queries and returns fresh, cited context.

LIVE TRACE

→ router.classify( ) · ok

→ researcher.fetch( ) · 212ms

→ responder.draft( ) · streaming…

LIVE TRACE

→ router.classify( ) · ok

→ researcher.fetch( ) · 212ms

→ responder.draft( ) · streaming…

FEATURES

The canvas is the monitor.

Execution theatre

Live node states, pulsing handoff edges, and streaming output turn every run into an observable system.

Planner

gpt-5 · reasoning

Researcher

tool · running

Web Search

fetch · 3 results

Debugger discipline

Pause, resume, step, breakpoints, and full run traces.

Local-to-cloud parity

One TypeScript engine runs your workflow in-app and in a Cloudflare Durable Object. Byte-identical. Deploy in one click.

Zero-setup local runs

Build and watch a real multi-agent run with no API key and no signup, powered by local models via Ollama.

Bring your own everything

Anthropic, OpenAI, Gemini, OpenRouter, Ollama. Your keys, your Cloudflare account. No lock-in.

The MCP flywheel

Any Model Context Protocol server becomes a draggable tool node instantly. Integrations grow with the whole ecosystem.

As fast as code. As simple as no-code. Built for agents, not steps.

n8n can’t debug agents like an IDE. LangGraph can’t be handed to your ops team. Phorma does both.

Phorma

Visual agent graph plus IDE-grade debugging.

Ops-readable canvas, developer-trustable traces.

Same engine local, desktop, and cloud.

Designed around agents, tools, and handoffs.

Integration tools

n8n, Zapier

Great for deterministic business automations, but agent runs need memory, tool calls, trace playback, and debugging discipline.

Flowchart-first. Not agent-debugger-first.

Code-first SDKs

LangGraph, CrewAI

Powerful for engineers, but hard to hand to operations teams without custom dashboards, replay tooling, and visual affordances.

Engineer-first. Not team-observable by default.

DEPLOY

From desktop studio to your cloud without a rewrite.

Prototype locally, watch every run, then push the exact same workflow engine into a Cloudflare Durable Object.

Desktop Studio

local run

Router

agent · gpt-5

Researcher

agent · running

trace.run() → streaming

one click, same engine

Cloudflare Durable Object

deployed

Same TypeScript engine

Byte-identical runtime. Your account. Your keys. No lock-in.

How it works

From sketch to shipped in three moves

01

Design the flow

Drag agents, tools, and guardrails onto the canvas. Wire them together like you think — visually.

02

Debug it live

Step through every call, inspect payloads, and replay failures with full traces — no print statements.

03

Deploy anywhere

One click to ship the exact graph you tested — to our cloud, your VPC, or your own metal.

Features

Everything an agent team needs

✓ plan.generate — 88ms

✓ tools.search — 212ms

⚠ guard.pii — retried ×1

✓ respond.stream — live

Time-travel debugging

Every run is a full recording. Scrub back to any step, edit the input, and replay from there.

Infinite canvas

60fps pan and zoom across thousands of nodes. Built like a design tool, not a dashboard.

Built-in evals

Score every change against golden datasets before it ever reaches production traffic.

Multiplayer by default

Live cursors, comments, and branch-based reviews for every workflow change.

Any model, any tool

Swap providers per node. Bring your own tools with a typed SDK or plain HTTP.

Why Phorma

Retire the notebook graveyard

Scripts, notebooks, and hope

Debugging via print statements and log greps

Orchestration logic scattered across five repos

“Works on my machine” becomes the deploy strategy

Nobody can see what the system actually did

Phorma Agent Studio

Full visual traces for every run, forever

One canvas — the source of truth for the whole team

What you test locally is exactly what ships

Watch agents think, call tools, and recover — live

Local to cloud

Build on your machine. Ship to any machine.

The Studio runs as a native desktop app with a local runtime, so you can iterate offline against real data. When it’s ready, promote the exact same graph to Phorma Cloud, your VPC, or self-hosted infrastructure — no rewrites, no config drift.

macOS · Windows · Linux

SOC 2 Type II

Self-host ready

phorma deploy

Live

$ phorma deploy --env production

✓ Graph validated — 14 nodes, 3 agents

✓ Evals passed — 98.4% on golden set

✓ Deployed to phorma.cloud in 6.2s

→ Streaming live traffic…

Trusted by agent teams at

Pricing

Start free. Scale when it works.

Free

$0

The full canvas, local runtime, and 1,000 cloud runs a month. For individuals.

Pro

Popular

$29

Per editor, monthly. Unlimited runs, evals, multiplayer, and priority routing.

Enterprise

Custom

VPC or self-hosted deploys, SSO and SCIM, audit logs, and a dedicated solutions engineer.

Ship agents you can actually see

Download the Studio and run your first multi-agent workflow in under five minutes.