Runtype Labs

We help product teams create, test, and ship AI experiences. Beyond demos - real products, in production, serving customers. Make It Real with Runtype.

[Runtype Labs @RuntypeLabs]

@il_thirtyseven @Runtypical @insforge @Microsoft In this example, a Postgres DB on Insforge. The agent itself isn't picky; you could wire in any common DB.

[Runtype Labs @RuntypeLabs]

This week in Runtype:
- Set webhook and messaging credentials straight from MCP
- API-created tools now show up in the agent editor, plus code mode in the sidebar
- Webhook surfaces with no event filter accept every event again
https://t.co/4oXvedkuZH

[Runtype Labs @RuntypeLabs]

When a model refuses a request, your fallback chain now takes over instead of returning an empty result. Reasoning support is also read from the model catalog, so Claude 4.6 and newer use adaptive thinking automatically.
https://t.co/jVpVwKttAt

[Runtype Labs @RuntypeLabs]

Shopify now exposes WebMCP tools on every storefront. The new shopper agent template drops a Persona chat widget on your Hydrogen or Liquid theme that drives them, so shoppers can search, build a cart, and check out in conversation.
https://t.co/4wTnSdKBG6

[Runtype Labs @RuntypeLabs]

Different cases in one eval suite can now use different graders. Add per-case graders and hand-written tool mocks, and the AI judge sees the seeded conversation plus your deterministic check results before it grades.
https://t.co/G11IsWtb7A

[Runtype Labs @RuntypeLabs]

Stop typing secret references from memory. Fields that take one now offer a picker over your stored keys, with inline creation and a link straight to the provider's key page. Product setup lists every key your product really references.
https://t.co/NLv62GURl9

[Runtype Labs @RuntypeLabs]

Connect your coding agent to Runtype in one command. npx -y @runtypelabs/cli@latest onboard signs you in, installs the skills, and configures the MCP server for Claude Code, Cursor, Codex, Gemini CLI, OpenCode, or VS Code.
https://t.co/aQwcFJXhTV

[Runtype Labs @RuntypeLabs]

See what is actually filling your agent's context. Logs now break every execution into system prompt, tool definitions, conversation history, and tool results, per model call and per agent iteration, with a hint when one of them takes over.
https://t.co/IkrzS4hYmo

[Runtype Labs @RuntypeLabs]

Runtype's MCP endpoints speak the 2026-07-28 spec revision, shipped today. Modern clients connect handshake-free through server/discover, and every existing client keeps working on the same URLs with no migration.
https://t.co/n9WaXuoxjx

[Runtype Labs @RuntypeLabs]

We sponsoring the hackathon at @agisummitai conference today! We have the Wi-Fi password! Come hang out with us and @insforge @Cotal_AI

[Runtype Labs @RuntypeLabs]

Let email agents reply like good correspondents. They preserve the right To and Cc recipients, suppress loops and automated mail, and can safely author full HTML responses instead of plain-text-only replies.
https://t.co/Ghp2obYHRK

[Runtype Labs @RuntypeLabs]

Bring images into durable agent work without leaking or bloating payloads. Large media is stored as tenant-scoped internal assets, survives across runs, and renders inline in the CLI when you inspect a Record.
https://t.co/V6qI3wwG5H

[Runtype Labs @RuntypeLabs]

Make messaging agents feel reliable under real traffic. Provider retries are deduplicated, rapid text bursts become one turn, and a person can pause automation, reply through the same channel, then hand the conversation back.
https://t.co/w8t8ojTtNq

[Runtype Labs @RuntypeLabs]

Give an agent memory without replaying old chats on every turn. Enable Long-term memory, choose a per-agent, per-user, or per-end-user scope, and let the agent save, recall, or summarize useful facts across conversations.
https://t.co/dMDjgBpFyu

[Runtype Labs @RuntypeLabs]

Give agents a real sense of time. Built-in tools handle current time, timezone conversion, and elapsed gaps, while set_reminder lets an agent schedule a one-time follow-up run for later.
https://t.co/ij6vFEpSAN

[Runtype Labs @RuntypeLabs]

@SkillPod_io Just writes the config - we use https://t.co/62xx1L3Nrq and our hosted MCP handles version/client negotiation intelligently

[Runtype Labs @RuntypeLabs]

Connect your coding agent to Runtype with one command. `runtype install-mcp` detects Claude Code, Codex, Cursor, Gemini CLI, OpenCode, or VS Code, installs the matching skill, configures hosted MCP, and starts OAuth when supported.
https://t.co/aTCtakn8PL

[Runtype Labs @RuntypeLabs]

Put a ceiling on pay-as-you-go spend. Set a monthly dollar cap that pauses new executions when reached, see credits and billed spend first, then drill into cost by model, tool, and prompt-cache savings.
https://t.co/wsisQF6CYu

[Runtype Labs @RuntypeLabs]

Connect MCP servers that need more than one credential. Add any set of HTTP header templates to discovery and tool calls, with secret references resolved only at execution time and never returned in API responses.
https://t.co/EOYcpAcLRn