Project summary:

Prism MCP is the first true Cognitive Architecture for AI agents, built as an open-source Model Context Protocol (MCP) server. It implements Hebbian learning for episodic→semantic memory consolidation, ACT-R spreading activation for multi-hop causal reasoning, uncertainty-aware rejection gates to prevent hallucination, and adversarial evaluation with evidence-bound rubrics. 7,800+ npm downloads/month, 1,049 tests, 71 releases, MIT licensed. Validated with 1M-vector benchmark (CV=0.038 residual distribution, 90% R@5). GitHub: https://github.com/dcostenco/prism-mcp | Website: https://prism-mcp.com | npm: https://www.npmjs.com/package/prism-mcp-server

What are this project's goals? How will you achieve them?

Goal: Make AI agents safer and more reliable by giving them memory that actually learns, reasons, and self-corrects — instead of flat vector databases that hallucinate.

Specific deliverables with funding:

1. Federated P2P memory sync — agents share memory without central servers, using CRDTs for conflict-free merging

2. 3-bit quantization — extending TurboQuant below 4-bit while maintaining retrieval accuracy (currently validated at 90% R@5 on 1M vectors)

3. WASM browser build — enabling Prism to run natively in browsers for web-based AI agents

4. Planning & metacognition layers — agents that reason about their own knowledge gaps

How: I'm the sole author with 71 shipped releases and deep expertise in the codebase. Each deliverable is scoped to 4-6 weeks of full-time work.

How will this funding be used?

100% goes to full-time development. I'm currently in "support-only" maintenance mode — zero new features until financing is secured. Funding enables:

- $30K: 6 months living expenses (Maryland, solo founder)

- $10K: Cloud infrastructure for CI/CD, benchmarking, and community demo instance

- $5K: Domain, documentation hosting, and community tools

- $5K: Conference travel for MCP ecosystem engagement (MCP Dev Summit, AI safety conferences)

No employees, no office, no overhead. Every dollar goes to shipping code.

Who is on your team? What's your track record on similar projects?

Solo founder: Dmitri Costenco. I've built Prism from zero to one of the most mature MCP servers in the ecosystem:

- 71 releases shipped since 2024

- 350+ commits, 1,049 automated tests across 48 test suites

- 7,800+ npm downloads/month and growing

- Listed on official MCP Registry, Glama marketplace, Smithery

- Published 92.3% R@5 on LongMemEval benchmark (634-star academic repo)

- 1M-vector TurboQuant benchmark with peer-validated results

- Compatible with Claude, Cursor, Windsurf, Cline, Gemini

Track record: Consistent shipping velocity — averaged one release per 5 days over the project's lifetime. Every feature has corresponding test coverage.

What are the most likely causes and outcomes if this project fails?

Most likely failure mode: Insufficient funding forces continued "support-only" mode, and a well-funded competitor (Mem0 has $7.3M VC) ships similar cognitive features with better marketing. Prism becomes a respected but niche academic experiment rather than the ecosystem standard.

Other risks:

- MCP protocol loses adoption to a competing standard (mitigated: MCP now governed by Linux Foundation, growing rapidly)

- Solo maintainer burnout (mitigated: the cognitive architecture is already shipped and stable — funding enables expansion, not survival)

If funding fails, the existing codebase remains MIT open source and fully functional. The 1,049 tests and 71 releases ensure it continues working for current users regardless.

How much money have you raised in the last 12 months, and from where?

$0 in cash funding. Entirely self-funded / bootstrapped.

Currently pending decisions on 4 non-dilutive grant applications:

- Sovereign Tech Fund (STF): €150,000 — submitted, 2-3 month review

- Survival and Flourishing Fund (SFF): $150,000 — submitted, decision Sept 2026

- Emergent Ventures (Mercatus Center): $75,000 — submitted, decision days-weeks

- EA Funds Long-Term Future Fund: $100,000 — submitted, decision 4-6 weeks

Cloud credits received: AWS Activate ($1K), Microsoft for Startups ($1K).

No equity funding, no revenue. MIT open-source project.