App 05 — Forge

Chain LLM calls into workflows.

DAG-based workflow engine. Chain LLM calls, transforms, and tool invocations with dependency ordering, retries, and full execution history.

Start free See the API

Run	Workflow	Steps	Duration	Status
run_a8f2	content-pipeline	5/5	8.7s	✓
run_7b31	support-triage	3/3	2.1s	✓
run_c4e9	data-enrichment	2/4	12.3s	✗
run_1d5a	content-pipeline	5/5	7.9s	✓

Run

Workflow

Steps

Duration

Status

run_a8f2

content-pipeline

5/5

8.7s

✓

run_7b31

support-triage

3/3

2.1s

✓

run_c4e9

data-enrichment

2/4

12.3s

✗

run_1d5a

content-pipeline

5/5

7.9s

✓

What Forge does

DAG Workflows

Define multi-step pipelines as directed acyclic graphs. Each node can be an LLM call, a tool invocation, or a data transform. Dependencies are resolved automatically.

Tool Registry

Register HTTP endpoints, transforms, and LLM calls as reusable tools. Reference them by name in any workflow. Add custom tools with a single POST.

Execution Tracking

Every workflow run is logged step by step. See which nodes completed, which failed, input/output at each stage, and total cost and latency.

Stateful Sessions

Maintain conversation context across workflow runs. Sessions track messages, tool calls, and state so multi-turn workflows work naturally.

Batch Processing

Submit arrays of inputs to run a workflow in parallel. Track batch progress and collect results. Ideal for bulk processing and data enrichment.

Visual Builder

Drag-and-drop workflow editor at /forge/builder/. Connect nodes with bezier edges, configure each step, and export the workflow JSON.

The API

Workflows, runs, tools, sessions, and batch jobs are all REST resources. Build complex LLM pipelines with simple HTTP calls.

# Create a workflow curl -X POST /api/forge/workflows \ -d '{"slug":"content-pipeline", "steps":[ {"id":"extract","type":"llm","model":"gpt-4o","prompt":"Extract topics: {{input}}"}, {"id":"sources","type":"tool","tool":"fetch_url","depends_on":["extract"]}, {"id":"draft","type":"llm","model":"gpt-4o","depends_on":["extract"]}, {"id":"edit","type":"llm","depends_on":["draft","sources"]}, {"id":"format","type":"transform","tool":"markdown_fmt","depends_on":["edit"]} ]}' # Run a workflow curl -X POST /api/forge/workflows/content-pipeline/run \ -d '{"input":"Latest developments in AI safety research"}' # Check run status and step details curl /api/forge/runs/run_a8f2/steps # Register a custom tool curl -X POST /api/forge/tools \ -d '{"name":"search_api","type":"http", "url":"https://api.search.io/v1/query","method":"POST"}' # Submit a batch curl -X POST /api/forge/batch \ -d '{"workflow":"content-pipeline", "inputs":["Topic A","Topic B","Topic C"]}'