RelayPlane MCP Server

Give AI coding agents access to multi-step AI workflows with 90%+ context reduction.

The Problem

When AI agents use MCP tools directly, they face two problems:

1.Tool definitions bloat context — 100k+ tokens for many tools
2.Intermediate results pass through context — a 50k token document copied between steps

The Solution

The RelayPlane MCP Server lets agents orchestrate AI workflows where intermediate results stay in the workflow engine, not your context window.

1# Without RelayPlane: 50k transcript flows through context twice
2TOOL CALL: gdrive.getDocument() → [50k tokens in context]
3TOOL CALL: salesforce.update({ notes: [50k tokens written again] })
4
5# With RelayPlane: transcript stays in workflow
6relay_workflow_run({
7  steps: [
8    { name: "fetch", mcp: "gdrive:getDocument", params: { id: "abc123" } },
9    { name: "save", mcp: "salesforce:update",
10      params: { notes: "{{steps.fetch.content}}" },
11      depends: ["fetch"] }
12  ]
13})
14# Only final result enters context

Result: 90%+ context reduction on multi-step pipelines

Quick Install

1claude mcp add relayplane -- npx @relayplane/mcp-server

Available Tools

Tool	Purpose	Cost
relay_run	Execute a single AI model call	Provider cost
relay_workflow_run	Execute multi-step workflows	Provider cost
relay_workflow_validate	Validate DAG structure (no LLM calls)	Free
relay_models_list	List available models with pricing	Free
relay_skills_list	Discover pre-built workflow patterns	Free
relay_runs_list	View recent execution history	Free
relay_run_get	Get run details and trace URL	Free

BYOK Pricing: Budget tracking protects you from runaway provider costs (OpenAI, Anthropic bills). RelayPlane is BYOK (Bring Your Own Keys) — we do not charge for API usage.

Supported Models

•OpenAI: gpt-5.2 (New!), gpt-5.2-pro, gpt-5-mini, gpt-5-nano, gpt-5.1, gpt-4.1, o3, o4-mini
•Anthropic: claude-opus-4.5, claude-sonnet-4.5, claude-haiku-4.5, claude-opus-4.1, claude-sonnet-4, claude-3.7-sonnet
•Google: gemini-3-pro, gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite
•xAI: grok-4, grok-4-fast, grok-3, grok-3-mini
•Perplexity: sonar-pro, sonar, sonar-reasoning-pro, sonar-deep-research
•Local: llama3.3, qwen2.5, deepseek-r1, mistral (via Ollama)

Next Steps

Installation — Setup for Claude Code, Cursor, and other clients
Tools Reference — Complete tool documentation with schemas
Skills — Pre-built workflow patterns with context reduction metrics
Budget & Limits — Configure safety limits for provider costs