Runtype

Runtype is a platform for shipping AI products: agents, flows, tools, surfaces, records, schedules, evals, and product templates. Use this skill as the entry point when the user is asking about Runtype broadly or needs help connecting an agent to the platform.

This skill is intentionally a router. For implementation work, switch to the narrower skill that matches the job.

First Move

If the Runtype MCP server is available, use it as the live source of truth before giving schema, catalog, or creation guidance:

• get_build_instructions(task="explain-capabilities") for scoping.
• get_build_instructions(task="build-product") before product construction.
• get_build_instructions(task="generate-flow") before flow construction.
• get_platform_documentation(topic=...) for schemas, surface traits, tool catalogs,

SDK docs, Persona embed docs, dashboard links, and type definitions.

If MCP is not connected, use the CLI-to-MCP golden path instead of presenting several equivalent setup choices:

1. Authenticate the CLI with runtype auth login so this session can keep working.
2. Run runtype install-mcp. It installs this skill, configures the current harness for

https://api.runtype.com/v1/mcp/protocol, and starts client-owned OAuth when possible.

1. Continue the current session with CLI commands. Tell the user to restart or reload the

harness because a running agent may not discover a newly configured MCP connection.

1. In the next session, use MCP first and call get_build_instructions before building.

Do not choose an API key merely to avoid a restart. Headless API-key setup is an exception for environments where browser OAuth is genuinely impossible; follow https://runtype.com/auth.md only after confirming that constraint with the user. The full public setup script is at https://runtype.ai.

Route To Focused Skills

Use this routing table instead of loading every Runtype detail into context:

| User intent | Use | | ----------------------------------------------------------------------------------------------------------------------------------------- | ----------------------- | | Build, deploy, or validate a product with agents, flows, tools, surfaces, records, secrets, schedules, or evals | runtype-build-product | | Inspect or modify a live account, debug failures, read logs/traces, compare evals, manage resources | runtype-admin | | Embed or theme a Persona chat widget, build fullscreen assistant layouts, use client tokens, or configure WebMCP/browser-side local tools | runtype-persona | | Package a product as a distributable FPO template, handle pending secrets, validate import readiness | runtype-templates | | Use the TypeScript/Python SDK, CLI, Marathon, playbooks, sandboxes, or code-first stored/upsert/virtual workflows | runtype-sdk-marathon |

Mental Model

• Product: the container the user ships.
• Agent: an LLM with a system prompt and tools. Start here for most interactive products.
• Flow: a deterministic pipeline. Use for fixed sequences, indexing, batch work, and hot

paths that should be fast and cheap.

• Tool: a typed callable, from built-ins, Orthogonal APIs, MCP servers, external HTTP,

custom code, local SDK tools, flows, or subagents.

• Surface: where users or machines reach the product, including chat, api, mcp,

mcp_code, webhook, email, slack, schedule, sms, imessage, discord, whatsapp, telegram, messaging, a2a, and hosted-page.

• Record: Runtype state and memory. It is not a replacement for the user's business

database.

• Eval: surface-level or capability-level comparison. Prefer surface/product evals when

measuring user experience.

• FPO template: the portable distribution format for a product.

When a product has multiple capabilities on one conversational surface, Runtype can provision an orchestrator that routes each incoming message to the right capability. Treat surfaces and capabilities as many-to-many.

Durable References

These local references are fallback context only. Prefer live MCP docs when available.

• references/primitives.md for the deeper platform model.
• references/mcp-tools.md for MCP tool groups and selection.
• references/surfaces.md for surface traits and prompt implications.
• references/flow-steps.md for flow step categories and common mistakes.
• references/persona-widget.md for Persona embed details.
• references/fpo-templates.md for template packaging.
• references/working-modes.md for dashboard, MCP, REST, SDK, and on-prem tradeoffs.
• references/recipes.md for worked product shapes.

Do Not Do

• Do not invent payload shapes. Fetch live docs or validate first.
• Do not let this router grow into platform docs. If more detail is needed, fetch

get_platform_documentation(topic="platform-catalog") or a focused topic, then read the relevant direct MCP resources when the docs point to them.

• Do not inline secret values. Use {{secret:KEY}} and pending-secret intake.
• Do not store all customer data in records; keep the source of truth where it belongs.
• Do not use Runtype for generic one-off LLM chat unless the user wants to ship a

repeatable product or operational workflow.