Update TODO: Add MCP as enabling technology for features

TODO.md

@@ -6,50 +6,54 @@ This list outlines the top 10 priority features identified to make this system a
     *   Implement database storage (e.g., SQLite, PostgreSQL) for research goals, context memory (hypotheses, results), run history, and user feedback.
     *   *Why:* Essential for any non-trivial use; allows resuming runs, comparing results across sessions, and managing generated data effectively.
 
-2.  **Enhanced Novelty/Feasibility Validation:**
-    *   Integrate external data sources for validation. Examples:
-        *   PubMed API search for novelty checks against existing literature.
+2.  **Leverage Model Context Protocol (MCP):**
+    *   Implement MCP servers to provide tools for external interactions (web search, database lookups, API calls).
+    *   Refactor agents to use `use_mcp_tool` for accessing external data/functionality needed for validation, RAG, market analysis, etc.
+    *   *Why:* Provides a standardized, robust, and extensible way to connect the AI to real-time external information and capabilities, enabling many other features on this list.
+
+3.  **Enhanced Novelty/Feasibility Validation (via MCP):**
+    *   Build MCP tools that wrap external APIs/services for validation:
+        *   PubMed API search tool for novelty checks.
         *   Patent database search (e.g., Google Patents API) for IP checks.
         *   Use specialized LLMs or structured data for deeper technical feasibility analysis.
-    *   *Why:* Grounds LLM assessments in real-world data, crucial for scientific rigor and entrepreneurial due diligence.
+    *   *Why:* Grounds LLM assessments in real-world data, crucial for scientific rigor and entrepreneurial due diligence. (Enabled by MCP).
 
-3.  **Market/Impact Assessment Agent:**
-    *   Add a dedicated agent or enhance the Reflection agent to evaluate hypotheses based on potential market size, societal impact, commercial viability, or alignment with business goals.
-    *   *Why:* Directly addresses entrepreneurial needs, adding a crucial dimension beyond purely technical assessment.
+4.  **Market/Impact Assessment Agent (via MCP):**
+    *   Add an agent or enhance Reflection agent to use MCP tools (e.g., web search, financial data APIs) to assess market size, impact, viability.
+    *   *Why:* Directly addresses entrepreneurial needs, adding a crucial dimension beyond purely technical assessment. (Enabled by MCP).
 
-4.  **User Feedback Integration & Weighted Ranking:**
+5.  **User Feedback Integration & Weighted Ranking:**
     *   Allow users to rate, tag, or comment on hypotheses within the UI during/after cycles.
     *   Modify the ranking system (e.g., Elo or a new multi-objective approach) to incorporate user feedback as a weighted factor alongside novelty, feasibility, etc.
     *   *Why:* Makes the user an active participant, tailoring results to their domain expertise and specific priorities.
 
-5.  **Experimental Design / Next Step Suggestion Agent:**
-    *   Add an agent that analyzes top-ranked hypotheses and suggests concrete next steps.
-    *   Examples: Propose specific experiments, identify necessary resources/expertise, suggest validation methods.
-    *   *Why:* Increases the actionability of the generated hypotheses, bridging the gap between idea and practical execution for scientists.
+6.  **Experimental Design / Next Step Suggestion Agent (via MCP):**
+    *   Add an agent using MCP tools (e.g., querying protocol databases, chemical suppliers) to suggest concrete next steps, experiments, resources.
+    *   *Why:* Increases the actionability of the generated hypotheses, bridging the gap between idea and practical execution for scientists. (Potentially enabled by MCP).
 
-6.  **Literature Corpus Integration (RAG):**
+7.  **Literature Corpus Integration (RAG via MCP):**
     *   Allow users to upload relevant documents (PDFs) or specify research areas/keywords.
-    *   Implement Retrieval-Augmented Generation (RAG) to provide this specific corpus as context to the LLM during generation and reflection steps.
-    *   *Why:* Massively improves the quality, relevance, and grounding of hypotheses within a specific domain.
+    *   Implement RAG, potentially using an MCP server/tool to interface with a vector database or document retrieval system containing the user-provided corpus.
+    *   *Why:* Massively improves the quality, relevance, and grounding of hypotheses within a specific domain. (Enabled by MCP).
 
-7.  **Advanced Visualization & Interaction:**
+8.  **Advanced Visualization & Interaction:**
     *   Enhance the similarity graph (e.g., node clustering, filtering by score/tags, sizing nodes by Elo).
     *   Add plots showing score/metric trends over iterations.
     *   Allow direct interaction with visualizations (e.g., selecting nodes to prune/evolve).
     *   *Why:* Improves usability, facilitates pattern recognition, and allows for more intuitive exploration of the hypothesis space.
 
-8.  **Structured Input & Constraints:**
+9.  **Structured Input & Constraints:**
     *   Develop a more structured way for users to define the research goal, including key variables, target outcomes, specific technical/budgetary constraints, target user/market segments.
     *   Use this structured input to generate more targeted prompts for agents.
     *   *Why:* Provides finer-grained control over the process, leading to more relevant and constrained results.
 
-9.  **Advanced Evolution Strategies:**
+10. **Advanced Evolution Strategies:**
     *   Implement more sophisticated methods for evolving hypotheses beyond simple combination.
     *   Examples: LLM-driven mutation based on critiques, varied crossover techniques, targeted refinement prompts.
     *   *Why:* Improves the core refinement capability of the system, potentially leading to more creative and robust hypotheses.
 
-10. **Run Comparison & Trend Analysis:**
-    *   Store results from multiple runs (potentially with different settings).
-    *   Add UI features to compare results across runs (e.g., which settings produced higher-ranked hypotheses?).
-    *   Visualize trends over time (e.g., average Elo score per cycle).
-    *   *Why:* Enables meta-analysis of the system's performance and helps users understand how different parameters influence the outcomes.
+11. **Run Comparison & Trend Analysis:**
+    *   Requires persistent storage (#1). Store results from multiple runs.
+    *   Add UI features to compare results across runs.
+    *   Visualize trends over time.
+    *   *Why:* Enables meta-analysis and understanding of how settings impact outcomes.