Build a mental models framework with a multi-agent dev team
A solo founder used Claude, ChatGPT, Windsurf and Cursor as role-based team members to design, test and deploy a 120-model mental-models framework and MCP server. Task: A solo founder coordinated Claude, ChatGPT, Windsurf, and Cursor as role-based team members using a SITREP protocol for parallel execution. They applied the framework's own mental models to architect, validate, and deploy the system, utilizing a 5-question wickedness rubric to tier problems. The human role involved defining these protocols and overseeing the multi-agent coordination, though specific manual coding tasks are not detailed.; tools: Claude, ChatGPT, Cursor. Human role: Not stated in the source; limitations: Results are unverified and rely on the author's self-assessment metrics. The 22-month timeline suggests high latency for solo multi-agent workflows. Integration requires specific tool access (Claude, ChatGPT, Windsurf, Cursor) and a custom SITREP coordination layer. Human oversight is critical for validating the 'meta-recursive' architecture logic..
Tools
Published on Agent Work Atlas:
Overview
A solo founder used Claude, ChatGPT, Windsurf and Cursor as role-based team members to design, test and deploy a 120-model mental-models framework and MCP server.
Workflow
A solo founder coordinated Claude, ChatGPT, Windsurf, and Cursor as role-based team members using a SITREP protocol for parallel execution. They applied the framework's own mental models to architect, validate, and deploy the system, utilizing a 5-question wickedness rubric to tier problems. The human role involved defining these protocols and overseeing the multi-agent coordination, though specific manual coding tasks are not detailed.
Reported result
The project produced 120 mental models with a reported 9.2/10 quality score and passed 140 chaos tests. An MCP server for Claude Desktop was deployed, and the entire effort took 22 months. These outcomes are self-reported by the author and lack independent verification.
Human role
Not stated in the source
Limitations
Results are unverified and rely on the author's self-assessment metrics. The 22-month timeline suggests high latency for solo multi-agent workflows. Integration requires specific tool access (Claude, ChatGPT, Windsurf, Cursor) and a custom SITREP coordination layer. Human oversight is critical for validating the 'meta-recursive' architecture logic.
Try it
Act as a technical lead managing a multi-agent development team. I want to build a small subset of a mental models framework using Claude, ChatGPT, and Cursor. Inputs: - Target: [Replace with specific problem domain, e.g., 'Startup Strategy'] - Models to generate: [Replace with list, e.g., 'First Principles, Inversion, Second-Order Thinking'] - Output Location: [Replace with repo path or file structure] Steps: 1. Define distinct roles for each AI agent (e.g., Claude for architecture, ChatGPT for content generation, Cursor for code implementation). 2. Establish a coordination protocol (SITREP style) to manage parallel tasks. 3. Generate the initial architecture and model definitions. 4. Implement the code using the assigned tools. 5. Run validation tests on the generated models. Deliverables: - A structured JSON file containing the model definitions. - A TypeScript module implementing the framework. - A test report showing pass/fail status for validation checks. Quality Check: - Ensure logical consistency between the architecture and implementation. - Verify that each model maps to exactly one transformation type. Safeguards: - Do not deploy to production. - Require human review of the architecture before coding begins. - If an agent fails to follow the role definition, pause and request clarification. Start by proposing the role definitions and coordination protocol for my review.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.