[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f2Rkj6eyep-B4_reQMusJF-38U6TN3DzXJ8VbB5WoQNk":3},{"item":4},{"id":5,"idKnowledge":6,"idDomain":7,"idCluster":7,"kindOverride":7,"slug":8,"title":9,"description":10,"bodyMarkdown":11,"bodyHtml":12,"author":13,"date":14,"createdAt":15,"topics":16,"image":19,"hasDownload":20,"fileName":7,"youtubeId":21},"81","35D1B267-BD8E-B545-850A-1A0A35CB8050","","filemaker-2026-ai-gets-serious-llms-gemini-support-and-agentic-development-on-the-horizon","FileMaker 2026: AI Gets Serious – LLMs, Gemini Support, and Agentic Development on the Horizon","Claris FileMaker 2026 (v26) brings major AI upgrades including Google Gemini support, smarter schema annotations for LLMs, and a roadmap toward fully agentic FileMaker development.","## FileMaker 2026: AI Gets Serious\n\nClaris officially released **FileMaker 2026** (internal version 26) in June 2026, \nand it's the most AI-forward release in the platform's 40-year history.\n\n### What's New for AI & LLMs\n\n**Google Gemini support**  \nFileMaker now supports Google Gemini alongside existing providers — OpenAI, Anthropic (Claude), \nand Cohere. Gemini also allows developers to pass container-field images directly into prompts \nfor visual analysis workflows.\n\n**Field & Table Annotations**  \nA long-standing frustration has been solved: you can now add a dedicated **field annotation** \nvia the Database Manager's Advanced options. This gives LLMs proper context about what a field \n*means* — not just what it's named. The new `FieldAnnotation()` and `BaseTableComment()` \ncalculation functions make this metadata accessible at runtime too.\n\n**Faster JSON & API calls**  \nWhen `Insert from URL` retrieves a JSON response, it is now automatically parsed and cached — \nmaking all subsequent JSON operations dramatically faster. A major win for solutions that \nmake heavy use of REST APIs and AI calls.\n\n**AI Model Server improvements**  \nThe AI Model Server now has its own version number and can be upgraded independently of \nFileMaker Server. Multiple Python processes mean embedding and text-generation workloads \nrun concurrently rather than queuing.\n\n**Timeout control**  \nText generation steps now accept a `CURLOPT_TIMEOUT` parameter in seconds — so if a \nlocal\u002Fopen-source model server stalls, your scripts no longer hang indefinitely.\n\n---\n\n### What's Coming: Agentic FileMaker Development\n\nThe bigger news may be what's *next*. Claris CEO Ryan McCann has outlined a vision for \n**agentic coding capabilities** — where tools like Claude Code, Cursor, and Codex can read \nyour FileMaker schema and automatically generate native scripts, tables, fields, layouts, \nand calculations.\n\nDeveloper previews are expected **later this summer (2026)**.\n\n---\n\n### Why This Matters for Your FileMaker Solutions\n\n- You can now connect FileMaker to **multiple AI providers** and choose based on cost, \n  capability, or data residency requirements.\n- **Schema annotations** make natural language queries and semantic search significantly \n  more accurate — without restructuring your database.\n- **Local LLM support** remains fully intact: run models like Llama 3 on-premise, \n  keeping sensitive data entirely within your own infrastructure.\n\nFileMaker 2026 treats AI as a **foundational platform layer**, not an add-on — \nand the agentic capabilities coming this summer could fundamentally change how \nFileMaker solutions are built.","\u003Ch2>FileMaker 2026: AI Gets Serious\u003C\u002Fh2>\n\u003Cp>Claris officially released \u003Cstrong>FileMaker 2026\u003C\u002Fstrong> (internal version 26) in June 2026, \nand it's the most AI-forward release in the platform's 40-year history.\u003C\u002Fp>\n\u003Ch3>What's New for AI & LLMs\u003C\u002Fh3>\n\u003Cp>\u003Cstrong>Google Gemini support\u003C\u002Fstrong>\u003Cbr>FileMaker now supports Google Gemini alongside existing providers — OpenAI, Anthropic (Claude), \nand Cohere. Gemini also allows developers to pass container-field images directly into prompts \nfor visual analysis workflows.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>Field & Table Annotations\u003C\u002Fstrong>\u003Cbr>A long-standing frustration has been solved: you can now add a dedicated \u003Cstrong>field annotation\u003C\u002Fstrong> \nvia the Database Manager's Advanced options. This gives LLMs proper context about what a field \n\u003Cem>means\u003C\u002Fem> — not just what it's named. The new \u003Ccode>FieldAnnotation()\u003C\u002Fcode> and \u003Ccode>BaseTableComment()\u003C\u002Fcode> \ncalculation functions make this metadata accessible at runtime too.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>Faster JSON & API calls\u003C\u002Fstrong>\u003Cbr>When \u003Ccode>Insert from URL\u003C\u002Fcode> retrieves a JSON response, it is now automatically parsed and cached — \nmaking all subsequent JSON operations dramatically faster. A major win for solutions that \nmake heavy use of REST APIs and AI calls.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>AI Model Server improvements\u003C\u002Fstrong>\u003Cbr>The AI Model Server now has its own version number and can be upgraded independently of \nFileMaker Server. Multiple Python processes mean embedding and text-generation workloads \nrun concurrently rather than queuing.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>Timeout control\u003C\u002Fstrong>\u003Cbr>Text generation steps now accept a \u003Ccode>CURLOPT_TIMEOUT\u003C\u002Fcode> parameter in seconds — so if a \nlocal\u002Fopen-source model server stalls, your scripts no longer hang indefinitely.\u003C\u002Fp>\n\u003Chr>\n\u003Ch3>What's Coming: Agentic FileMaker Development\u003C\u002Fh3>\n\u003Cp>The bigger news may be what's \u003Cem>next\u003C\u002Fem>. Claris CEO Ryan McCann has outlined a vision for \n\u003Cstrong>agentic coding capabilities\u003C\u002Fstrong> — where tools like Claude Code, Cursor, and Codex can read \nyour FileMaker schema and automatically generate native scripts, tables, fields, layouts, \nand calculations.\u003C\u002Fp>\n\u003Cp>Developer previews are expected \u003Cstrong>later this summer (2026)\u003C\u002Fstrong>.\u003C\u002Fp>\n\u003Chr>\n\u003Ch3>Why This Matters for Your FileMaker Solutions\u003C\u002Fh3>\n\u003Cul>\n\u003Cli>You can now connect FileMaker to \u003Cstrong>multiple AI providers\u003C\u002Fstrong> and choose based on cost, \ncapability, or data residency requirements.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Schema annotations\u003C\u002Fstrong> make natural language queries and semantic search significantly \nmore accurate — without restructuring your database.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Local LLM support\u003C\u002Fstrong> remains fully intact: run models like Llama 3 on-premise, \nkeeping sensitive data entirely within your own infrastructure.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>FileMaker 2026 treats AI as a \u003Cstrong>foundational platform layer\u003C\u002Fstrong>, not an add-on — \nand the agentic capabilities coming this summer could fundamentally change how \nFileMaker solutions are built.\u003C\u002Fp>\n","Shubham","2026-06-26",1782453679000,[17,18],"AI","FileMaker","\u002Fapi\u002Fknowledge\u002Fimage\u002F81\u002F?v=7a51f6e7d9bf",false,null]