CFFBRWCFFBRWDocumentation
GuideAPIMCPOpen Dashboard
⚠Auto-generated documentation — This content was generated with AI assistance and is currently under review. Information may be incomplete or inaccurate. When in doubt, contact our support team.

Getting Started with CFFBRW

Get up and running with CFFBRW in under 5 minutes.

What is CFFBRW?

CFFBRW is an AI-powered workflow platform where you describe automation in plain English markdown. An AI compiler translates your prose into a deterministic execution DAG, and Cloudflare edge infrastructure runs it with self-healing, live streaming, and universal tool integration via MCP.

One-liner: Write workflows in markdown. AI compiles them. Cloudflare runs them.

Quick Start

  1. Go to cffbrw.com and sign up
  2. Pick a template from the gallery — or start from blank
  3. Click Compile to see the AI-generated execution plan
  4. Click Run to watch the DAG execute live in your notebook
  5. Edit any step in markdown and recompile

Your first workflow can run in under 2 minutes.

Your First Workflow

Step 1: Create a new workflow

From the dashboard, click New Workflow → Start from Blank.

Step 2: Write markdown

# Weather Report

## Steps

### 1. Fetch weather data
Get the current weather for Tokyo from OpenWeather API.
The URL is https://api.openweathermap.org/data/2.5/weather?q=Tokyo&units=metric&appid={API_KEY}

### 2. AI summary
Describe the weather in one sentence. Include temperature and conditions.
This step depends on step 1.

Step 3: Compile

Click Compile. The AI reads your markdown and produces an execution plan with:

  • Step types (http_request, ai_call)
  • Dependencies (dependsOn)
  • Parallel execution batches

Step 4: Run

Click Run. Watch the DAG execute in real-time with step-level streaming output.

Key Concepts

The AI Compiler

The compiler reads your markdown and:

  1. Parses step descriptions and dependency hints ("depends on step X")
  2. Infers the execution graph (DAG) from your prose
  3. Classifies each step into a type (HTTP, AI, transform, etc.)
  4. Validates the plan with 8 heuristic checks
  5. Caches the plan by content hash — same markdown = zero AI cost

Execution Model

CFFBRW uses a three-tier execution model:

TierTypeBehavior
Tier 1DeterministicSteps run as planned (HTTP calls, transforms, code)
Tier 2AI RecoveryWhen Tier 1 fails, AI diagnoses the error and fixes it
Tier 3Plan EvolutionWhen AI recovery isn't enough, the plan itself is re-compiled

Step Types

Step TypeDescription
http_requestMake HTTP calls to any API
ai_callSend context to the AI for analysis or generation
transformManipulate data from previous steps
conditionalBranch execution based on a condition
loopIterate over a collection
waitPause for a duration or until a time
human_approvalPause and wait for human input
codeExecute JavaScript in a sandbox
mcp_callCall an MCP server tool
sub_workflowRun another workflow as a step

Cron Scheduling

Automate your workflows on a schedule:

  1. Open your workflow → Settings
  2. Set a cron expression (e.g., 0 9 * * 1-5 for weekdays at 9am UTC)
  3. Enable the schedule

Your workflow will run automatically. View past runs in the Run History tab.

MCP (Model Context Protocol)

MCP lets you integrate any API as a workflow tool without writing integration code. Register an MCP server in your dashboard, and reference its tools directly in your markdown steps.

See the MCP Integration Guide for details.

Next Steps