Mesh Mimir - Cardano AI Stack - Contextual RAG, ChatGPT, MCP

[Proposal setup] Proposal title

Please provide your proposal title

Mesh Mimir - Cardano AI Stack - Contextual RAG, ChatGPT, MCP

[Proposal Summary] Budget Information

Enter the amount of funding you are requesting in ADA

200000

[Proposal Summary] Time

Please specify how many months you expect your project to last

8

[Proposal Summary] Translation Information

Please indicate if your proposal has been auto-translated

No

Original Language

en

[Proposal Summary] Problem Statement

What is the problem you want to solve?

Cardano's high developer barrier is throttling dApp adoption and user onboarding. This immediate stagnation costs us key market share against competing chains.

[Proposal Summary] Supporting Documentation

Supporting links

• https://meshjs.dev/
,
• https://github.com/MeshJS/mimir

[Proposal Summary] Project Dependencies

Does your project have any dependencies on other organizations, technical or otherwise?

No

Describe any dependencies or write 'No dependencies'

No dependencies

[Proposal Summary] Project Open Source

Will your project's outputs be fully open source?

Yes

Please provide details on the intellectual property (IP) status of your project outputs, including whether they will be released as open source or retained under another licence.

Mesh Mimir is fully open source under a permissive license. All code, tools, and improvements are freely available for any Cardano project to use, modify, and distribute without restrictions.

[Theme Selection] Theme

Please choose the most relevant theme and tag related to the outcomes of your proposal

AI

[Campaign Category] Category Questions

Describe what makes your idea innovative compared to what has been previously launched in the market (whether by you or others).

Mesh Mimir is the first fully open-source AI tooling stack that works not only for Cardano projects but anywhere teams use MDX documentation. Unlike existing solutions, Mimir provides ready-to-deploy RAG pipeline, a custom ChatGPT, and an MCP server that stay always in sync with project documentation.

No months of engineering setup are needed. Teams just configure their repository, and Mimir handles ingestion, context generation, embeddings, and deployment automatically. Developers immediately get accurate, context-aware AI assistance from their own project docs, improving onboarding, productivity, and safety.

Mimir enables projects to adopt AI tooling instantly and collaboratively, transforming how developers interact with documentation. It also brings full flexibility in choosing LLM providers, whether open-source, self-hosted, or commercial, so teams can use the models that best match their performance, security, and cost requirements.

Describe what your prototype or MVP will demonstrate, and where it can be accessed.

The MVP demonstrates a fully functional AI tooling stack currently used for Mesh. It includes:

RAG pipeline for automatically ingesting project documentation.

Custom ChatGPT on project's website providing answers from the documentation.

MCP server for AI-assisted queries in code editors.

The proposal will abstract this system so any Cardano project can deploy it quickly using a single configuration file, enabling AI-powered tooling for their own documentation.

References:

• Mesh Website with AI features: https://meshjs.dev, https://meshjs.dev/ai
• Mesh MCP Server NPM Package: https://www.npmjs.com/package/meshjs-mcp

Describe realistic measures of success, ideally with on-chain metrics.

Adoption & ecosystem usage

• 10+ Cardano projects publicly deploy and integrate Mesh Mimir within 12 months
• Public testimonials from teams confirming faster developer onboarding

Dev-experience / efficiency indicators

• Reduction in time-to-first-successful-transaction for new developers working on participating projects (measured through internal dev surveys or onboarding cohorts)
• Reduction in repeated support queries in community channels (e.g., common “how do I…?” questions answered by the AI assistant instead of core contributors)

[Your Project and Solution] Solution

Please describe your proposed solution and how it addresses the problem

Every day, capable developers try Cardano and bounce off. The learning curve is too steep. Haskell, fragmented tooling, scattered documentation. They leave for easier chains, taking their dApps, users, and innovation with them.

This problem is far worse today because expectations have changed. With more than 33 million Copilot users and tools like Cursor becoming standard, everyone from developers to product managers now codes with AI at their side. They expect instant answers, guidance, and clarity as they build. If Cardano cannot meet that expectation, they assume the ecosystem is behind. To keep these builders, Cardano must become the first AI-powered blockchain environment, where intelligent assistance is the default experience rather than a luxury.

The solution is AI assistance, but building it takes months. A production Retrieval Augmented Generation (RAG) pipeline requires 2-3 months of engineering. Keeping it synced with evolving docs takes another month. Model Context Protocol (MCP) integration adds more time. Custom ChatGPT for websites keeps adding months. Most teams never start at all.

Mimir eliminates this timeline entirely. One config file deploys everything: RAG pipeline, MCP integration, and custom ChatGPT. What took months now takes minutes.

The core is an advanced RAG system with Contextual Retrieval. Standard RAG loses critical context when documents are split into chunks. Mimir adds an LLM preprocessing step that generates specific context for each chunk before embedding, ensuring developers get not just code snippets but full understanding of security implications and design patterns.

Setup requires only a config file. Provide your GitHub docs URL, Supabase credentials, and LLM API keys. Deploy once. The system handles everything: chunking, context generation, embedding creation, and index maintenance. GitHub webhooks keep it synced automatically using checksums, so only changed content gets re-embedded.

Once the RAG pipeline is live, configure your backend URL and publish the MCP server as a package to npm. It integrates directly into VSCode and Cursor, wrapping your project's functionality into tools that AI agents can call directly. Developers describe what they want in natural language, and the AI executes functions, queries state, and runs tests.

The same RAG infrastructure powers custom ChatGPT widgets on websites. Users get instant answers from your documentation, dramatically improving onboarding and reducing support tickets.

Most critically for web3, code written with MCP assistance follows current best practices from up-to-date documentation, not outdated training data. This is essential where a single mistake can mean lost funds. Mimir ensures developers get current, safe patterns, dramatically reducing vulnerability risks.

For Cardano projects, the cost of lacking this infrastructure is severe. Teams spend 4 to 6 months building their own RAG systems instead of delivering the DEX, NFT marketplace, or governance product that attracts users. New developers lose their first month just learning how to perform basic transactions instead of adding value. Talented builders who hit friction early simply leave, taking their dApp ideas, their users, and their momentum with them. Mimir changes the game. Deploy in an afternoon, redirect months of engineering effort back into real product development, enable contributors on day one, and retain the builders who would otherwise walk away.

Small projects get the same AI infrastructure as major companies. Solo developers get the same assistance as established teams. With Mimir barrier isn't lowered. It's removed.

[Your Project and Solution] Impact

Please define the positive impact your project will have on the wider Cardano community

Mimir directly addresses Cardano's biggest growth bottleneck: developer accessibility. Every project that deploys Mimir becomes a gateway that keeps developers in the ecosystem instead of losing them to complexity.

The immediate impact is measurable. Each team/project save 4-6 months of engineering time they would spend building AI infrastructure, redirecting that capacity to building actual dApps. New developers onboard in days instead of weeks, becoming productive contributors faster. Support teams stop drowning in basic documentation questions, freeing them to solve real problems.

The compounding impact is ecosystem-wide. As more projects deploy Mimir and improve their documentation for AI compatibility, the quality of Cardano's collective knowledge base rises. Better documentation means better AI assistance across all projects.

Developers who stay because of accessible tooling build the next generation of dApps. Projects that ship faster attract more users. Users who onboard smoothly bring more users.

This creates a virtuous cycle. Cardano's technical sophistication stops being a barrier and becomes accessible. The ecosystem that has the strongest technology gains the developer velocity it needs to compete. Projects that would have failed at onboarding succeed. Developers who would have left stay and build. Users stop leaving out of confusion and start actually using the product.

Most critically, Mimir proves that Cardano's complexity doesn't need to be simplified, just made accessible. This preserves what makes Cardano superior while eliminating what makes it intimidating. The result is an ecosystem that grows without compromising its technical foundations.

[Your Project and Solution] Capabilities & Feasibility

What is your capability to deliver your project with high levels of trust and accountability? How do you intend to validate if your approach is feasible?

Mesh has been a longtime contributor to the Cardano open-source developer ecosystem since 2022. Our tools are widely used, and we have proven our skills and commitment by building and maintaining essential infrastructure that empowers developers to build on Cardano. Over this time, through non-stop building on Cardano, we have gained expertise that positions us uniquely to deliver Mimir's AI infrastructure.

We are proud to have participated in Catalyst since Fund 10, and we are well familiar with the process of completing proposals. For full transparency on all our funded proposals, their progress and details, we have built a dashboard that makes it easy to assess and audit our capabilities: https://gov.meshjs.dev/catalyst-proposals. Our track record speaks to our commitment to "don't trust, verify."

Mimir is not theoretical. We have already built this entire infrastructure for Mesh itself. Mesh is now AI-ready with a full RAG pipeline using Contextual Retrieval, automatic GitHub webhook syncing, and MCP integration. We use it daily in production.

The challenge we solved for ourselves took months of engineering. This proposal is about abstracting that complexity away so any Cardano project can deploy the same system in minutes. The hard technical work is done. Now we are building the abstraction layer that makes it effortless for everyone else.

Our feasibility is proven by Mesh's production system. The technology works. This proposal funds making it accessible to the entire Cardano community.

[Milestones] Project Milestones

Milestone Title

Build Core RAG Pipeline

Milestone Outputs

In the first milestone, we build the complete RAG pipeline specialized for MDX documentation. This includes MDX parsing, chunking, embedding, vector indexing, and GitHub webhook syncing so documentation updates automatically feed into the RAG system.

• Complete ingestion pipeline including parsing, chunking, and embedding.
• GitHub webhook system for incremental updates (only re-ingest changed files/chunks).
• Query API returning accurate context.
• Published update post to inform the community.

Acceptance Criteria

• End-to-end ingestion working on at least two real Cardano projects.
• MDX parsing handles headings, code blocks, components, and frontmatter.
• Query API produces accurate responses for queries.
• Published X/Twitter update announcing the milestone.

Evidence of Completion

• Links to GitHub PRs containing the full RAG pipeline.
• Link to X/Twitter announcement post.
• Link to documentation

Delivery Month

2

Cost

40000

Progress

40 %

Milestone Title

Add LLM Flexibility & CLI Tooling

Milestone Outputs

This milestone focuses on making the system adaptable and developer-friendly. Teams will be able to plug in any LLM provider and run ingestion or maintenance tasks through a simple CLI.

• LLM provider abstraction layer supporting open-source, self-hosted, or commercial models.
• CLI tool for running ingestion, validating config, and refreshing docs.
• Published X/Twitter update to inform the public.

Acceptance Criteria

• Multiple LLM providers supported with no code changes.
• CLI validates configs and runs ingestion successfully on all major OS.
• Published X/Twitter update to inform the public.
• Link to documentation

Evidence of Completion

• GitHub PRs containing LLM abstraction layer and CLI implementation
• Link to X/Twitter announcement.
• Link to documentation.

Delivery Month

1

Cost

40000

Progress

50 %

Milestone Title

One-File Configuration and Deployment Services

Milestone Outputs

This milestone delivers a one-file configuration system and full deployment tooling so teams can launch their RAG infrastructure with a single command, including Docker images and platform templates.

• One-file config enabling zero-code setup (GitHub repo URL, Supabase credentials, LLM API keys).
• Docker images for API server, ingestion worker, webhook listener, and vector DB.
• docker-compose.yml for local and production deployments.
• Optional deployment templates for platforms like Render or Railway
• Published X/Twitter update.

Acceptance Criteria

• Full RAG system deploys successfully using docker compose up.
• All services auto-connect and load configuration.
• Logs clearly show MDX ingestion and document updates.
• Public announcement post published.

Evidence of Completion

• Links to GitHub PRs containing Dockerfiles and deployment templates.
• X/Twitter post demonstrating deployment.
• Link to documentation.

Delivery Month

1

Cost

40000

Progress

60 %

Milestone Title

Custom ChatGPT for Websites

Milestone Outputs

This milestone focuses on integrating the already-built RAG backend with a clean, production-ready chat UI that any project can embed on their website. The backend is complete from previous milestones; here we deliver a polished frontend layer + a simple integration flow.

• A polished, modern chat UI that connects to the existing streaming chat backend.
• Simple copy-paste integration snippet so teams can drop the chat UI into any website.
• Documentation explaining how to connect the UI to the backend using only the config file + API URL.

Acceptance Criteria

• Chat UI successfully connects to the existing RAG backend and streams responses.
• Fully working integration implemented on ecosystem project.
• Public milestone announcement published.
• Link to usage and documentation.

Evidence of Completion

• GitHub PR links for the frontend chat UI and integration docs.
• Live demo page showing the chat UI connected to the backend.
• Link to public announcement on X/Twitter.
• Link to usage and documentation.

Delivery Month

1

Cost

20000

Progress

70 %

Milestone Title

MCP Server for Code Editor Integration

Milestone Outputs

This milestone delivers the MCP server that brings documentation directly into developer workflows through tools like Cursor, Zed, Windsurf, and VSCode MCP clients.

• MCP server exposing tools for searching docs and generating examples.
• Published MCP server as an NPM package.
• Editor integration examples for common MCP-enabled editors.
• Published X/Twitter update.

Acceptance Criteria

• Developers can ask AI about the project’s docs directly inside their editor.
• The AI can successfully generate a complete, working code solution by using MCP Server tools.
• Public announcement post.

Evidence of Completion

• Links to GitHub PRs with MCP implementation.
• Published NPM package link.
• Demo video showing MCP functionality in editors.
• Link to X/Twitter post.

Delivery Month

2

Cost

35000

Progress

90 %

Milestone Title

Documentation & Onboarding

Milestone Outputs

In the final milestone, we focus on creating comprehensive documentation, guides, and onboarding resources to ensure that any Cardano project can deploy and use the system successfully. This is about making adoption effortless and helping teams get productive immediately.

• Full step-by-step guides for:
  • Configuring the MDX RAG pipeline
  • Deploying services via Docker
  • Setting up the Custom ChatGPT
  • Using MCP inside editors
• Troubleshooting guide for common MDX issues.
• Public announcement of the fully documented system via X/Twitter.

Acceptance Criteria

• Clear, complete documentation that allows a team to go from zero → deployed system without assistance.
• Onboarding guides include example configurations and usage workflows.
• Public announcement published

Evidence of Completion

• Links to full documentation and guides in GitHub repo
• Links to X/Twitter post announcing milestone completion

Delivery Month

1

Cost

25000

Progress

100 %

[Final Pitch] Budget & Costs

Please provide a cost breakdown of the proposed work and resources

1. Build Core RAG Pipeline — 50,000 ADA

Scope

• Design and implement the full MDX-specialized RAG pipeline:
  • MDX parsing (headings, components, code blocks, frontmatter)
  • Chunking strategy tuned for developer documentation
  • Embedding + vector indexing
  • GitHub webhook integration for incremental updates
• Implement query API that returns accurate, filtered context for LLMs.
• Deploy and validate on at least two real Cardano projects.

2. LLM Flexibility & CLI Tooling — 35,000 ADA

Scope

• LLM provider abstraction layer supporting open-source, self-hosted, or commercial models.
• CLI tool for running ingestion, validating config, and refreshing docs.

3. Deployment Services (Docker Images & Infrastructure) — 35,000 ADA

Scope

• One-file config enabling zero-code setup (GitHub repo URL, Supabase credentials, LLM API keys).
• Docker images for API server, ingestion worker, webhook listener, and vector DB.
• docker-compose.yml for local and production deployments.
• Optional deployment templates for platforms like Render or Railway

4. Custom ChatGPT for Websites — 20,000 ADA

Scope

• Build a polished, modern chat UI (web widget) that connects to the streaming RAG backend.
• Provide a copy-paste integration snippet so projects can embed the chat UI on any website.
• Document the integration flow using only config + API URL.
• Implement and demonstrate the UI on at least one ecosystem project.

5. MCP Server for Code Editor Integration — 35,000 ADA

Scope

• Implement an MCP server exposing tools for:
  • Searching docs.
  • Fetching examples.
  • Generating code snippets based on project documentation.
• Package and publish MCP server as an NPM module.
• Provide integration examples for MCP-enabled editors (Cursor, Zed, Windsurf, VSCode MCP clients).

6. Documentation & Onboarding — 25,000 ADA

Scope

• Produce full, end-to-end documentation and guides for:
  • Configuring the MDX RAG pipeline.
  • Deploying via Docker / deployment templates.
  • Adding the Custom ChatGPT to websites.
  • Using the MCP server in editors.
  • Troubleshooting common MDX / ingestion issues.
• Ensure a team can go from zero → fully deployed system without direct assistance.

[Final Pitch] Value for Money

How does the cost of the project represent value for the Cardano ecosystem?

Mesh has long been a cornerstone of Cardano development, powering hundreds of projects and widely adopted across the ecosystem. With Mesh Mimir, we are extending this impact into AI-assisted development. Mimir allows any team to deploy a production-ready RAG pipeline, a project-aware ChatGPT, and MCP integration in hours instead of months, which significantly reduces engineering time and cost.

Mimir is fully open-source, so every improvement benefits all Cardano projects immediately. By lowering the barrier to entry for AI-assisted development and keeping documentation and code in sync, Mimir helps teams build faster, safer, and smarter. Every ADA invested increases developer productivity, improves onboarding, and strengthens the ecosystem. This makes the project one of the highest-value investments for Cardano open-source development today.

[Self-Assessment] Self-Assessment Checklist

I confirm that evidence of prior research, whitepaper, design, or proof-of-concept is provided.

Yes

I confirm that the proposal includes ecosystem research and uses the findings to either (a) justify its uniqueness over existing solutions or (b) demonstrate the value of its novel approach.

Yes

I confirm that the proposal demonstrates technical capability via verifiable in-house talent or a confirmed development partner (GitHub, LinkedIn, portfolio, etc.)

Yes

I confirm that the proposer and all team members are in good standing with prior Catalyst projects.

Yes

I confirm that the proposal clearly defines the problem and the value of the on-chain utility.

Yes

I confirm that the primary goal of the proposal is a working prototype deployed on at least a Cardano testnet.

Yes

I confirm that the proposal outlines a credible and clear technical plan and architecture.

Yes

I confirm that the budget and timeline (≤ 12 months) are realistic for the proposed work.

Yes

I confirm that the proposal includes a community engagement and feedback plan to amplify prototype adoption with the Cardano ecosystem.

Yes

I confirm that the budget is for future development only; excludes retroactive funding, incentives, giveaways, re-granting, or sub-treasuries.

Yes

[Required Acknowledgements] Consent & Confirmation

I Agree

Yes