Skip to content

Firefly Weave

Learn and use Firefly Weave

Start small: create a workflow, see how it runs, then connect it to your systems and your people.

Firefly Weave is a durable workflow orchestration and integration platform with human tasks. You describe a business process once, as a workflow. Weave runs each case of it, waits for people and events, calls your systems, and keeps the history of every step. You can draw workflows in Studio, write them as YAML, or manage them from the CLI, the Python SDK, or your own product.

New to Weave? Read Start here. It explains the four pieces of the platform and gives you an ordered path for your role. Coming from a BPM suite? Coming from BPM/BPMN translates BPMN tasks, gateways, and events into Weave steps.

Choose your path

Each path is an ordered list of guides; Start here gives the steps and what you finish with.

You are… You want to… Start with Then
A process designer or business analyst Draw processes, add approvals and API calls, and run them Concepts and your first workflow Studio, a local platform or your team's, the step reference, human tasks, and REST calls from a step
An integration developer Call APIs and systems, and build connectors and workers Connect the CLI to a platform Workflow authoring, publish and run with the CLI, REST calls without code, the SDK, a custom integration, and workers
An administrator Install the platform, configure sign-in, and give people access Start a local platform Identity and secrets, remote deployment, people and access, and configuration
An operator Keep runs healthy, fix incidents, upgrade, and back up Connect the CLI, then manage runs Incidents, troubleshooting, observability, upgrades, backup, and debug-session retention

Start each path by installing the CLI from the v0.1.0a14 release.

Who does what explains these roles with an order-approval example, and which Weave permissions each one needs.

Files, decisions, and AI

  • Work with files: upload documents, pass references through workflows, and transfer content from workers.
  • Decision tables: put reusable business rules behind a versioned decision step.
  • Run AI steps: configure workflow model profiles and operate an Agentic worker.
  • Configure Lumi: give Studio its own assistant configuration, separately from workflow AI.

Choose your next task

Your goal Start here You will finish with…
Try a workflow without running any service Install the CLI → first workflow A successful local simulation
Draw and edit a workflow Studio A visual editor that works locally and publishes to a connected platform
Start your own platform Start a local platform An API, database, identity service, a saved run, and a person who can sign in
Connect to a platform your team runs Connect the CLI or connect Studio A saved platform with your sign-in and chosen workspace
Configure each kind of step Studio step reference The right properties, expressions, and bindings for each step
Ask a person for a decision Human tasks An assigned, durable approval task
Call a REST API from a step Call a REST API without code An Action on the built-in HTTP connector, run from a workflow
Build your own integration Build a custom integration A custom action or an incoming webhook workflow
Implement and run a task handler Worker walkthrough → worker deployment An admitted handler that claims and completes work
Add Weave to a Python product Python SDK tutorial A workflow published and run from Python
Try API requests in a browser API playground Swagger connected to your platform
Process an email conversation Email Incoming messages and replies linked to a thread
Use your organization's identity provider Identity provider setup Verified tokens, published sign-in settings, identity links, and scoped grants
Give people access People and access People linked to their accounts, with scoped Weave roles
Deploy on your own infrastructure Remote deployment An ordered path through AWS, Azure, or Google Cloud

The first three things to understand

  1. Installing the CLI gives you a client. It can write, validate, compile, and simulate workflows without any service, open Studio, and call a running platform.
  2. Starting a platform gives you durable runs. The API keeps workflows and runs in PostgreSQL and accepts people who sign in through your identity provider, as long as Weave grants them access.
  3. Integrations connect external systems. The built-in HTTP connector calls JSON APIs with no code; your own workers run custom integration code. A Transform step needs neither.

How a request moves through clients, the API, storage, and workers

Follow the numbered boxes from a client's request through the API and the database to the integration code that calls your systems. Open diagram at full size

Read a tutorial one step at a time

Run one command block, check its Expected result, then continue. Comments inside code blocks explain why each command is there. When a terminal shows server logs, keep it open and use a second terminal for client commands.

The guides pin the v0.1.0a14 alpha release, a preview, which includes every feature they describe, such as saved platforms, REST calls without code, and the Studio editor. When a guide shows how to work with an alpha6 or earlier client or server, it says so. Capabilities and limits separates what is implemented, what was checked locally, and what you must still verify with your own providers.

Find an exact interface

Interface Reference
Every API operation, request, response, and schema Full API reference
Authentication, errors, revisions, and idempotency API behavior
CLI commands and file formats CLI reference
Python clients and builders SDK reference
Workflow language and expressions Definition contracts and compiler
Worker execution Worker protocol
Studio editor extension points, test data and Execute step Studio editor contracts
Server and client settings Configuration

The API and SDKs tab holds the browser playground, the Python tutorial, every HTTP operation and schema, and the SDK reference. You can browse the public reference without installing anything; Swagger on your running platform lets you send authorized requests to it.

Go deeper after the first successful run