The Local-First AI
Workflow Engine
Orchestrate multi-step AI workflows with any model. Run locally with zero latency or deploy to cloud. Full observability. No gateway tax.
QUICKSTART
Runs in 30 seconds with your own provider keys. Zero config BYOK.
One API for every AI model
Route between OpenAI, Anthropic, Google, xAI, and Ollama in seconds.
What Changes When You Use RelayPlane
| The Old Way | The RelayPlane Way |
|---|---|
| Different SDK for each provider | ✓One SDK, every provider |
| Agents generate inconsistent patterns | ✓Copy-paste examples that just work |
| Magic model resolution breaks silently | ✓Explicit provider:model — what you write is what runs |
| DIY retry, fallback, caching | ✓Built-in reliability |
| Observability is an afterthought | ✓Telemetry from day one |
| Tool calls need separate plumbing | ✓MCP steps native in workflows |
Production Templates
Pre-built workflows ready to deploy
90% Less Context, 100% Verified Code
The RelayPlane MCP Server lets Claude Code, Cursor, and other AI agents orchestrate multi-step workflows with massive context savings.
ONE-LINE INSTALL
Intermediate results stay in the workflow engine, not your context window
Run workflows, list models, discover skills, view traces
OpenAI, Anthropic, Google, xAI with unified interface
| Tool | Purpose | Cost |
|---|---|---|
| relay_workflow_run | Execute multi-step workflows | Provider cost |
| relay_run | Single model call | Provider cost |
| relay_workflow_validate | Validate DAG structure | Free |
| relay_skills_list | Discover pre-built patterns | Free |
| relay_models_list | List available models | Free |
Why RelayPlane Works
Built on principles that eliminate common AI workflow bottlenecks
No network overhead
Local execution means zero latency between steps. Your workflow runs at the speed of your machine, not your internet connection.
No API gateway bottleneck
Direct connections to providers eliminate the middleman. No queue, no rate limits beyond the provider's own.
Direct provider connections
Your API keys talk directly to OpenAI, Anthropic, Google, or xAI. No proxy, no logging, no security concerns.
Type-safe from end to end
Full TypeScript support with Zod schemas. Catch errors at compile time, not runtime. IntelliSense guides you through every step.
Simple, Transparent Pricing
No usage fees. No gateway tax. Pay for features, not tokens.
- ✓100 runs/month
- ✓1M tokens/month
- ✓1 cloud workflow
- ✓Community support
- ✓1,000 runs/month
- ✓10M tokens/month
- ✓10 cloud workflows
- ✓Run history & logs
- ✓Email support
- ✓10,000 runs/month
- ✓100M tokens/month
- ✓Unlimited cloud workflows
- ✓Webhooks + schedules
- ✓Team access (5 seats)
- ✓50,000 runs/month
- ✓500M tokens/month
- ✓High-frequency schedules
- ✓Team access (20 seats)
- ✓Priority support
- ✓Unlimited runs & tokens
- ✓SAML SSO
- ✓VPC deployment
- ✓Unlimited seats
- ✓Dedicated support
Frequently Asked Questions
How does BYOK work?
Bring Your Own Keys means we never proxy or store your API keys unless you choose to encrypt them in your dashboard. Security by design, not as an add-on. You maintain direct relationships with AI providers like OpenAI, Anthropic, and Google, with complete transparency and no markup costs.
Do you store my API keys?
On the Free plan, keys are passed inline or via environment variables—we never store them. On paid plans (Starter and above), you can optionally store keys in our dashboard, encrypted with AES-256. You can always use inline keys instead of dashboard storage.
How does pricing work?
RelayPlane pricing is simple and transparent. Free gives you 100 runs/month. Starter ($29/month) adds 1,000 runs and cloud telemetry. Pro ($99/month) unlocks webhooks, schedules, and team access. Scale ($399/month) adds high-frequency schedules and priority support. Enterprise gets custom SLAs.
What is MCP?
MCP (Model Context Protocol) lets you call external tools as workflow steps. Connect to CRMs, GitHub, Slack, or any MCP-compatible server. Mix AI steps with API calls in a single DAG workflow.
Do I need MCP to use RelayPlane?
No! MCP is optional. Most workflows only use AI steps. MCP is useful when you need to integrate external tools like CRMs, databases, or APIs into your workflow.
Which providers are supported?
RelayPlane supports OpenAI (including GPT-4o vision), Anthropic (Claude), Google (Gemini), xAI (Grok), and local LLMs via Ollama. All providers use the same unified API. Switch providers with a single line of code.
What happens if workflows fail?
RelayPlane includes built-in exponential backoff with jitter. Configure retry logic per workflow step. Handle rate limits and transient failures automatically. All error states are tracked and logged for debugging.
Start Building Today
No gateway tax. No lock-in. Full control. Join developers building the future of AI workflows.