Stop buying tokens.
Start shipping multilingual campaigns.qualified leads.
Production-ready AI workflows that replace token spend with defined outcomes, inside your security boundary.
</> EditDelete
Click a node to inspect its config
Kimi K2.5131K context
Qwen3 235B Instruct262K context
gpt-oss 120B131K context
Qwen3 Coder262K context
+ Your model$ graphn model deploy
Managed inference · OpenAI-compatible API
Start from a working blueprint
Pick a blueprint, save, and publish. No blank canvas.
Explore blueprints →Media
Production Studio
Turn any talk or podcast into a launch kit — chapters, highlight reel, show notes, and dubbed teasers in six languages
ingesttranscribeedit_plantranslate ×6cut_reelshow_notesdubpackage
.claude/skills/graphn-workflow/SKILL.md
.cursor/rules/graphn.md
.graphn/skills/workflow-building.md
.graphn/skills/dsl-schema.md
.graphn/skills/dsl-commands.md
.graphn/skills/debugging.md
Build and iterate from your terminal
graphn init writes skills for Claude Code and Cursor, so your coding agent can build and edit workflows with you.
Read the docs →Test, run, and trace every execution
Dry-run drafts before publishing, then inspect step-level traces and the full history of every run.
See how runs work →</> EditDelete
Input ⌘ Resources
Runs the published workflow. Use the editor to test changes before publishing.
{
"media_url": "https://images-assets
.nasa.gov/…/WeAreGoing~mobile.mp4"
}History
019fdabc11:42:36 AM
019fdaab11:31:02 AM
019fda9d11:24:48 AM
Output
—
Trace
ingestfunction
transcribefunction
edit_planagent
translate ×6parallel
cut_reelfunction
show_notesagent
dubfunction
packagefunction
</> EditDelete
syncasyncbatchevent
curl -X POST \
Publish once, call it anywhere
Every published workflow is a versioned API with sync, async, batch, and event gateways.
API reference →Built-in inference, or bring your own model
Serve open-weight models from the built-in catalog, import your own with the Python SDK, or deploy custom inference.
About models →</> EditDelete
Models
Built-in, imported, and custom-deployed models for your agents
Kimi K2.5
kimi-k2-5
Frontier open-weight hybrid-thinking model for agents, reasoning, and tool use
Qwen3 235B Instruct
qwen3-235b
Largest open-weight MoE model (22B active) for complex instruction following, agents, and long-context tasks
gpt-oss 120B
gpt-oss-120b
Open-weight Apache-2.0 MoE reasoning model with adjustable reasoning depth
Qwen3 Coder
qwen3-coder
Code-specialized MoE model for agentic software engineering and repo-scale context
import graphn
Replace token spend with defined outcomes.
SupportSupport RouterRoute customer messages to the right specialist for orders, returns, or product questionsEngineeringAutomated PR Review BotClone a repo, analyze a PR with a code-specialized model, synthesize review comments, and post themSearch & KnowledgeCompany Knowledge SearchAsk natural-language questions and get cited answers from your internal docsFinanceInvestment Due DiligenceAnalyze financials and market data to produce a Go/No-Go investment memoData & AnalyticsNatural Language SQLType a question, get a validated SQL query and formatted resultsHR & PeopleResume ScreenerUpload resumes and a job description, get a ranked shortlist with match reasoning
Browse all blueprints →Everything an agent needs
Chat with your agents
Talk to your published workflows in plain chat.
Learn more →Turn this keynote into a launch kit: images-assets.nasa.gov/…/WeAreGoing~mobile.mp4
Connect to Claude & ChatGPT
Published workflows show up as tools in any MCP host. One URL connects them all.
About Connect AI →ClaudeChatGPTCursor
https://gateway.graphn.ai/v1/<workspace>/mcpPublished workflows appear as tools in any MCP host.
production_studio
city_explorer
get_workflow_run
› localize this keynote for our six launch markets
One workspace, every resource
Agents, models, functions, tools, knowledge bases, storage, and secrets in one place.
Core concepts →Agents
Models
Functions
Tools
Knowledge Base
Storage
Secrets