Graph Compose: Hosted Temporal Workflows with Visual Builder and AI

Graph Compose is a hosted platform for orchestrating API workflows on Temporal's durable execution engine. You define workflows as graphs of nodes (HTTP calls, AI agents, iterators, error boundaries) that run as Temporal workflows without managing infrastructure.
Three Ways to Build the Same Graph
Every workflow uses a consistent JSON format that can be created through:
- A React Flow visual builder
- A typed TypeScript SDK (@graph-compose/client)
- An AI assistant that turns plain English into a graph
How It Works
You submit JSON workflows via API to https://api.graphcompose.io/v1/workflows/execute and they run immediately on hosted Temporal. No deploy step, no Temporal cluster to run, no workers to deploy. The platform handles scheduling, retries, error boundaries, and state persistence.
Workflow Example
const workflow = {
"nodes": [
{
"id": "upload_image",
"type": "http",
"dependencies": [],
"http": {
"method": "PUT",
"url": "https://storage.example.com/{{ context.filename }}",
"headers": {
"Content-Type": "{{ context.contentType }}",
"Authorization": "Bearer {{ $secret('storage_token') }}"
},
"body": "{{ context.imageData }}"
},
"activityConfig": {
"retryPolicy": {
"maximumAttempts": 3,
"initialInterval": "1s",
"backoffCoefficient": 2
},
"startToCloseTimeout": "30 seconds"
}
},
{
"id": "generate_thumbnails",
"type": "http",
"dependencies": ["upload_image"],
"http": {
"method": "POST",
"url": "https://images.example.com/process",
"body": {
"source": "{{ results.upload_image.data.url }}",
"sizes": ["400x400", "200x200"]
}
}
}
],
"context": {
"filename": "profile.jpg",
"contentType": "image/jpeg"
}
}
Key Features
- Durable Execution: Workflows survive server restarts, network failures, and process crashes. Temporal replays from the last checkpoint.
- Automatic Retries: Configure retry policies per node with max attempts, backoff intervals, and non-retryable error types.
- Error Boundaries: Wrap nodes in try/catch-style error boundaries with fallback behavior.
- State Persistence: Query workflow state at any point.
- Template Expressions: Nodes reference results from upstream nodes via
{{ results.node_id.data.field }}syntax.
Licensing
Open-core model: the execution foundations and integrations service are AGPL-3.0, while the platform orchestrator, visual builder, and AI assistant are proprietary.
📖 Read the full source: HN AI Agents
👀 See Also

Three MCP servers for e-commerce research with Claude: Shopify, Amazon, and Google Maps tools
A developer built three MCP servers for Claude to analyze Shopify stores without API keys, score Amazon product opportunities, and find/scored local business leads from Google Maps. All are available on Apify.

Send OpenClaw Agents as Meeting Participants with Voice, Chat, and Screen Share
A new skill lets OpenClaw agents join Google Meet, Teams, and Zoom as full participants with voice (STT/TTS), chat, video avatar, and screen sharing.

Skill Bill: A Markdown-Based Governance Framework for AI Coding Skills
A developer built Skill Bill, a framework of 44 Markdown-based AI skills for Kotlin, Android/KMP, PHP, and Go that addresses prompt management problems like naming drift and duplicated logic. It includes orchestrator skills like 'feature-implement' that chain 10-12 skill invocations and syncs to Claude Code, Copilot, GLM, and Codex.

skill-depot: A Local-First Memory and Skill System for MCP-Compatible AI Agents
skill-depot is a retrieval system that stores agent knowledge as Markdown files and uses vector embeddings to semantically search and selectively load only relevant content. It runs 100% locally with no API keys, works with any MCP-compatible agent, and can be set up with npx skill-depot init.