Make AI Agent Updates: New Models, Fallback Connections & MCP Tools

The short answer

Make's latest release touches the parts of the platform builders care about most: you can now improve an agent's response with Maia, set a fallback AI connection, enable prompt caching, and call two newer models — Claude Sonnet 5.5 and GPT-6.1 Sol. Enterprise teams also get SSO social-login routing and MCP tools for audit logs, all without moving to a higher tier^[1].

Maia can now improve your agent responses

Maia, Make's built-in AI assistant, can now review a scenario run together with you, tell you what to improve, and propose a revised version you can choose to apply^[1][2].

• Improve from a scenario run — start from the response your agent actually produced.
• Run-aware guidance — Maia reviews the run, the tools used, and the response to spot where changes would help.
• Targeted questions — tell Maia what to prioritize, like adding a missing step or tightening the final response.
• Versioned improvements — review the proposed version before you apply it.

Fallback connections & prompt caching

Make AI Agents now support a fallback AI provider connection: when the primary provider or model is unavailable, Make retries the run once using the fallback — with Make's own AI Provider as the default^[1].

Prompt caching is also new, letting an agent reuse instructions and tool definitions to cut latency and credit usage. You can select a caching duration for supported Anthropic Claude models (cached context applies across repeated calls in the same conversation), while supported OpenAI models automatically reuse prompt content^[1].

Two new models

Claude Sonnet 5.5

Now available in the Make AI Toolkit, Anthropic Claude, and Make AI Agent (New) apps, Claude Sonnet 5.5 brings a context window of up to 1,000,000 tokens and text outputs of up to 128,000 tokens — useful for processing large text collections and drafting long materials^[1].

GPT-6.1 Sol

GPT-6.1 Sol is also live in Make AI Toolkit, OpenAI (ChatGPT, Whisper), and Make AI Agent (New). It supports Functions, Web search, File search, and Computer use, with multi-agent support in beta for coordinating subtasks into a single implementation plan^[1].

Auto-route social logins to SSO

Organizations with a verified domain and SSO configured can now route social sign-ins into their SSO flow. Users starting with Google, Apple, GitHub, or Facebook on a verified domain are directed to the org's SSO with their email pre-filled — keeping IT control while preserving familiar sign-in choices^[1].

MCP tools for audit logs

Admins can now query audit-log events through MCP — listing events, filtering by user, action, or date, and inspecting individual entries — so scenario changes and organization activity can be surfaced inside the agents and apps teams already use. Access is read-only^[1][3].

Frequently asked questions

Are these updates on all plans?

Yes. Make states that the Maia improvements, fallback connections, prompt caching, new models, SSO routing, and MCP audit-log tools are available on all plans^[1].

Which apps include Claude Sonnet 5.5 and GPT-6.1 Sol?

Both are available in Make AI Toolkit and Make AI Agent (New). Claude Sonnet 5.5 is also in the Anthropic Claude app, while GPT-6.1 Sol is also in the OpenAI (ChatGPT, Whisper) app^[1].

What does a fallback connection do?

If your agent's primary AI provider or model is unavailable, Make automatically retries the run once using a fallback connection — with Make's AI Provider as the default fallback^[1].