I don’t know too much about MCP or AI or how technology works, but couldn’t you build some tools or scripts to let the MCP perform some testing? Also couldn’t you create some skill docs that the MCP could have so it doesn’t take the first 20 minutes to figure stuff out?
By the way, when in March 2025 did you send the message about MCP? This is what AI gave me recently
I searched the Bubble forum for the earliest and most prominent advocates of Bubble providing MCP access so external AI agents could read, build, and edit Bubble applications.
There are two distinct groups: early advocates who requested the capability before Bubble announced it, and developers who subsequently demonstrated or pushed its adoption.
1. Earliest identifiable proponents
These are ordered by the earliest relevant public statement I could verify, not by influence.
Date
Forum user
What they advocated
Mar 25, 2025
mo.salama
Proposed MCP protocol support within Bubble’s AI roadmap, primarily for building AI-powered apps.
Mar 26, 2025
bartgroupany
Suggested MCP-based conversational development as part of the future of AI app-building platforms.
Jun 8, 2025
rjwilkinson10
Proposed integrating Figma’s MCP with Bubble’s AI builder so generated UI could follow established design systems.
Jul 23, 2025
Atomily
Asked Bubble to provide MCP server creation capabilities similar to n8n and Xano.
Aug 3, 2025
nznvlogs
Explicitly requested MCP capabilities to directly edit a Bubble application’s UI.
Aug 3, 2025
NoCodeROI_Agency
Supported official API access for interface manipulation and had already released a Bubble Data API MCP server.
Aug 4, 2025
boston85719
Discussed exposing the editor APIs so third parties could create their own editor interfaces.
Feb 20, 2026
code-escapee
Explicitly called for Bubble to provide an MCP with read/write access to application JSON instead of forcing developers to use workarounds.
Feb 21, 2026
boston85719
Proposed publicly exposing the editor APIs and treating Bubble’s infrastructure as its core business.
The earliest confirmed general MCP suggestion was mo.salama on March 25, 2025.
The earliest explicit proposal I found for an MCP to edit the Bubble UI was nznvlogs on August 3, 2025.
The strongest early proposal for read/write access to the app-building system itself was code-escapee on February 20, 2026, followed by boston85719 the next day.
2. Biggest proponents of editor-level MCP access
Based on the strength of their public advocacy, repeated arguments, and practical contributions, these are the most significant individuals I identified.
1. georgecollier — Most significant technical contributor
February 12, 2026
Released Buildprint 2.0 with MCP connectivity to external AI agents, then demonstrated AI-driven changes to Bubble applications. His work provided concrete evidence that third-party agentic development was possible without Bubble’s official editor MCP. The announcement thread grew into a major ongoing community discussion.
2. code-escapee — Most explicit early advocate for official read/write MCP
February 19, 2026
Created BUBS, an AI agent using Buildprint MCP and structured JSON write operations. On February 19, explicitly requested stable MCP-style endpoints for app structure. On February 20, called for direct read/write access to Bubble JSON so third-party tools could operate without workarounds.
3. boston85719 — Strongest infrastructure-first strategic argument
August 4, 2025 (early discussion); February 21, 2026 (clear proposal)
Argued that Bubble should publicly expose its editor APIs and let third parties build competing development interfaces. Repeatedly advocated repositioning Bubble around infrastructure rather than its visual editor, including in March 2026. In June, also requested MCP access to plugin development capabilities.
4. nznvlogs — Earliest explicit UI-editing MCP requester found
August 3, 2025
Asked for MCP to directly manipulate Bubble UI, comparing the time required to build visually in Bubble against AI development products like Lovable and Bolt.
5. pyli — Strongest explicit urgency argument found
June 4, 2026
Argued that Bubble needed a true MCP connection to external LLMs to remain competitive and criticized prioritizing Bubble’s own AI chat system over open connectivity.
6. rjwilkinson10 — Early design-system integration advocate
June 8, 2025
Advocated Figma MCP integration so Bubble AI could follow existing design systems. In February 2026, publicly identified Buildprint as the most exciting contemporary development for Bubble.
3. Other important MCP pioneers
These contributors mainly worked on MCP access to apps built with Bubble, which differs from MCP access to Bubble’s editor.
User
Earliest relevant contribution
NoCodeROI_Agency
Aug 2, 2025 — Released an open-source Bubble Data API MCP server; supported expanding it to UI editing.
branimir.ivanov
Oct 20, 2025 — Published a practical implementation of an MCP server using Bubble backend workflows.
design.agx
Dec 14, 2025 — Released a tool for creating MCP servers from Bubble apps and continued expanding its functionality.
mo.salama
Mar 25, 2025 — Made the earliest verified general MCP suggestion; launched FlowCP in June 2026.
4. Findings
For the specific idea behind Bubble’s October 6, 2026 editor MCP release, three contributions stand out:
- nznvlogs made the earliest verified explicit request for MCP to edit a Bubble interface, on August 3, 2025.
- code-escapee and boston85719 articulated the case for official access to Bubble’s internal editing capabilities, rather than making developers rely on unofficial methods.
- georgecollier provided the most substantial practical demonstration through Buildprint.
An additional finding is that code-escapee explicitly requested stable MCP-style endpoints on February 19, 2026, one day earlier than the February 20 post identified in the initial research.
Bubble subsequently confirmed on September 3, 2026 that it was developing an MCP allowing external agents to build and edit applications, before releasing the public beta on October 6, 2026.
These are evidence-based assessments, not a verified ranking of every forum member by post count, likes, or influence. I cannot establish from public posts whether any particular contributor directly influenced Bubble’s internal product decision.