# Edison Experimentation Prover MCP for AI Agents AI Agent Connect > Edison Experimentation Prover forces your AI client to stop guessing and start testing. Instead of letting an agent pick the first solution it finds, this Connector requires it to evaluate multiple alternatives, document failures, and prove viability before moving forward. It is built for high-stakes R&D where best practice is not enough and you need a systematic way to avoid expensive mistakes. ## Overview - **Category:** complex-reasoning - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_UMZPgT3WihcCcSB3Nac36KXLZfuGNzCN6Yn2jL8p/ai-agent-connect - **Tags:** systematic-experimentation, rapid-prototyping, iteration, product-viability, thomas-edison, failure-documentation, ecosystem-design ## Description When you are building a complex system, it is easy for an AI agent to just pick the first solution that works and call it a day. That is how you end up with a product that works in a vacuum but falls apart under real-world pressure. This Connector changes that by forcing the agent to adopt a systematic R&D methodology. It makes the agent act like a lead engineer who knows that the first idea is rarely the best one. You will see it start to look for at least three different approaches for every major decision, scoring them against actual metrics rather than just vibes. It won't let the agent skip over the boring but necessary stuff like monitoring plans, rollout strategies, and documentation for the paths that did not work. It is designed to catch those good enough moments where a solution is technically functional but operationally incomplete. By using this through the Vinkius catalog, you are essentially installing a rigorous peer-review process into your agent's brain. It ensures that every choice is backed by data, every failure is recorded as a learning point, and every final decision is actually ready for production. Instead of just getting a block of code, you get a validated strategy that accounts for cost, friction, and long-term maintenance. ## Tools ### validate_edison_experimentation Forces the agent to test three distinct approaches and document the root causes of every failure. It ensures every decision has a data-backed justification and a clear path for iteration. ## Prompt Examples **Prompt:** ``` Should we use a relational database or a graph database for our new social connection feature? ``` **Response:** ``` To make this decision, I will evaluate three options: PostgreSQL, Neo4j, and a managed NoSQL service. | Option | Throughput | Query Complexity | Cost | | --- | --- | --- | --- | | PostgreSQL | High | Medium | Low | | Neo4j | Medium | Low | Medium | | Managed NoSQL | Very High | High | High | **Recommendation:** Neo4j is best for deep relationship queries, but PostgreSQL with a JSONB index is our best balance of cost and performance for the first 6 months. ``` **Prompt:** ``` We need to move our customer records from Excel to a new system. What is the best way? ``` **Response:** ``` I have analyzed three migration paths: - **Manual Import:** Low cost, high error risk. - **Scripted ETL:** Medium cost, high reliability, requires 48 hours of dev time. - **Third-party Migration Tool:** High cost, instant results. **Winner:** Scripted ETL. It allows for validation at each step. I have also drafted a 6-day transition plan and a staff training guide. ``` **Prompt:** ``` I want to build a new notification system. Can we just use a simple webhook? ``` **Response:** ``` A simple webhook works for low volume, but we need to consider the grid. I have evaluated the webhook approach against a message queue (RabbitMQ). - **Webhook:** Easy to set up, no retry logic. - **Message Queue:** Harder to set up, handles spikes, provides 99.9% reliability. **Verdict:** We should use a message queue. I have documented the failure modes of webhooks (timeout, duplicate delivery) to inform our choice. ``` ## Capabilities ### Evaluate multiple alternatives The agent compares three distinct approaches against your specific project constraints and metrics. ### Document failure root causes Every dead end is recorded with a clear explanation of why it failed and what was learned. ### Verify operational viability The agent must prove success metrics, operating costs, and user friction before finalizing a design. ### Force prototype commitment The agent is required to test a solution at a small scale to eliminate risks before full commitment. ### Map iteration cycles It tracks specific metrics for each version to identify when improvements reach diminishing returns. ## Use Cases ### Choosing a database without testing A developer wants to use PostgreSQL because it is the standard. The agent uses the Connector to compare it against NoSQL and a managed service, documenting the trade-offs in throughput and cost. ### Rushing a feature launch A PM wants to launch a new billing module. The agent uses the Connector to ensure there is a rollback plan, a monitoring dashboard, and a migration path for existing users before it considers the task done. ### High failure rate in prototypes A team keeps hitting walls with a new API. The agent uses the Connector to analyze why the last three attempts failed and identifies the specific technical bottleneck to guide the next attempt. ### Scaling a manual process A company is moving from manual data entry to automation. The agent uses the Connector to test three automation workflows and selects the one with the lowest friction for the staff. ## Benefits - Stop shipping good enough code by forcing the agent to find the best possible solution through validate_edison_experimentation. This ensures your team does not just pick the first idea that works but actually evaluates the highest quality option for your specific constraints. - Avoid expensive migration costs by requiring a prototype at 10% scale before full commitment. This Connector forces the agent to identify specific risks in a small environment so you do not waste your entire budget on a 100% rollout that fails in production. - Build complete ecosystems instead of isolated features by verifying rollout and monitoring plans. It ensures your agent does not just build a tool but also creates the grid of documentation, training guides, and recovery plans needed for a real product. - Turn failed attempts into an asset by documenting root causes and learnings for every dead end. Instead of just saying an idea did not work, the agent records why it failed and how that knowledge informs the next iteration, preventing repeated mistakes. - Get clear viability metrics on operating costs and user friction before you write a single line of production code. This forces the agent to provide a cost projection and success metrics before you commit to a design, ensuring commercial viability. ## How It Works The bottom line is that it turns your agent into a rigorous peer-reviewer for every major technical choice. 1. Give your AI client a high-level goal or a specific design decision to make. 2. The Connector prompts the agent to list alternatives and run a validation check. 3. You get a structured report showing why a winner was chosen and what was learned from the losers. ## Frequently Asked Questions **What is the Edison Experimentation Prover MCP?** It is a tool that forces your AI to think like a systematic researcher. It prevents the agent from picking the easiest solution and makes it test multiple options before committing to a design. **How does this help with expensive tech mistakes?** It requires the agent to document dead ends. By understanding why an idea failed before you spend money on it, you avoid repeating those same mistakes in your production environment. **Can I use this to choose a tech stack?** Yes, that is a primary use case. The Connector ensures the AI compares at least three different technologies against your specific constraints like cost, speed, and scalability. **Does it help with product rollout plans?** Absolutely. It won't let the agent consider a feature complete unless it includes a plan for monitoring, user onboarding, and transition steps. **Is this for simple coding tasks?** No, this is for complex reasoning. If you just need a quick function, you do not need it. Use it when you need to justify a major architectural choice or a high-stakes business decision. **How does it ensure my solution is viable?** It forces the agent to define success metrics and operating costs before it starts building. This gives you real-world evidence that the solution will work under pressure. **Doesn't requiring experiments slow down decision-making?** A 2-day pilot testing 3 alternatives is faster than an 8-month correction after choosing wrong. Edison's Menlo Park produced a major invention every 10 days — systematic experimentation is FASTER than guessing, because you avoid expensive dead ends. The team that picked the traditional system without testing spent 6 weeks migrating. A 3-day comparison pilot would have cost 1/30th of the migration. **What counts as sufficient iteration?** Test variations until you hit diminishing returns. At minimum 3 variations beyond the first working version. Measure the delta between each: V1 → V2 improved throughput 40%. V2 → V3 improved 12%. V3 → V4 improved 2%. Diminishing returns at V3 — ship V3. Edison tested 10,000 battery experiments but he was measuring progress. If V2 shows < 3% improvement over V1 for a non-critical metric, iteration on THAT axis is done. Move to the next constraint. **How does it differ from the Watt Efficiency Prover?** Watt validates OPTIMIZATION of existing systems — measuring waste, isolating bottlenecks, quantifying improvement. Edison validates DECISION-MAKING for new choices — testing alternatives, iterating designs, building ecosystems, proving viability. Watt asks 'where is 80% of your resources being wasted?' Edison asks 'how many alternatives did you test before choosing this one?' Use Watt when tuning a running operation. Use Edison when making a strategic choice or building a new capability.