Open Grant Proposal: Aiviber Persistent AI Agent Memory on Filecoin

Project Name: Aiviber

Proposal Category: Developer and Data Tooling

Individual or Entity Name: Nodefleet LLC

Proposer: nodefleet

Project Repo(s): https://github.com/nodefleet/aiviber

(Optional) Technical Sponsor: N/A

Do you agree to open source all work you do on behalf of this RFP under the MIT/Apache-2 dual-license?: Yes

Project Summary

Aiviber is a persistent AI agent memory platform that stores team knowledge and agent context on Filecoin, making every agent session smarter than the last. AI agents today lose all context when a session ends architectural decisions, domain knowledge, and learned corrections disappear and must be rebuilt from scratch every time. Aiviber solves this by maintaining a shared Company AI Brain backed by Filecoin storage, accessible to every agent and teammate via MCP, with cryptographic guarantees that the knowledge persists permanently.

Impact

Pain points this project addresses:

Every team using AI agents faces the same problem: agents are stateless. A developer agent that learned a smart contract's architecture last week has forgotten it today. Teams compensate by manually re-briefing agents every session, wasting time and producing inconsistent results.

The deeper problem is storage. Teams building their own memory systems default to centralized databases with no cryptographic proof of persistence and no open standard others can build on.

Aiviber solves both. Persistent agent memory backed by Filecoin for verifiable, permanent storage. The longer a team uses it, the more capable their agents become.

Benefits of getting this right:

Aiviber establishes Filecoin as the storage layer for AI agent intelligence: structured knowledge that compounds over time with high retrieval frequency. For developers, agents remember what they learned and share knowledge across the team. For the Filecoin ecosystem, a new category of high-value data written and read continuously.

Risks of not getting this right:

Without this, teams build fragmented centralized memory systems with no verifiability. The Filecoin ecosystem misses the AI agent memory vertical as it becomes one of the fastest-growing categories in developer tooling.

Impact:

Success means Aiviber becomes the standard memory layer for AI agents in Web3. Every team that adopts it becomes a continuous Filecoin data contributor. Impact is measurable in total knowledge stored, active agent sessions, and teams with a live Company AI Brain.

Outcomes

Primary deliverable: A production-ready AI agent memory platform backed by Filecoin, accessible via MCP to Claude, Cursor, VS Code, and Windsurf, with four memory modes and an open source SDK.

The software will deliver:

Company AI Brain on Filecoin: a shared knowledge store backed by Filecoin deals with permanent, verifiable persistence. CIDs surface in the dashboard for every knowledge entry. Shared across every agent and teammate simultaneously.

Four Memory Modes: Session Memory with configurable TTL, Project Memory scoped to a codebase or protocol, Team Knowledge Base for cross-project intelligence, and Skill Repository for versioned agent capabilities.

MCP-Native Integration: Aiviber connects to Claude, Cursor, VS Code, and Windsurf via MCP. Agents read and write Filecoin-backed memory automatically during normal operation.

Filecoin Deal Engine: automated deal-making for all knowledge writes, minimum three storage providers per entry, deal status and CIDs surfaced in dashboard.

Memory Dashboard: real-time view of the Company AI Brain, Filecoin deal status, CIDs, verification, and retrieval analytics.

Open Source SDK: released under MIT/Apache-2 for self-hosted Aiviber deployments on Filecoin.

Success metrics:
Total knowledge stored on Filecoin (GB)
Active Filecoin deals backing agent memory
Teams with a live Company AI Brain
MCP-connected agent sessions using Filecoin-backed memory
Retrieval latency for cached knowledge (target: under 100ms)
Teams self-hosting via the open source SDK

Data Onboarding

AI agent memory is a new category of Filecoin data: structured, frequently written, high retrieval frequency, and permanently valuable. The following projections are based on observed knowledge generation rates from Nodefleet's own agent operations.

Month 1: 50 GB
Internal deployment. Nodefleet's own development and infrastructure teams use Aiviber in production. Agent sessions across protocol operations, RPC infrastructure management, and AI tooling generate the initial knowledge corpus.

Month 3: 500 GB
Public beta. First external teams onboarded: blockchain development teams, protocol operators, and AI agent developers. Multiple Company AI Brains active across different organizations, each generating continuous knowledge writes.

Month 6: 5 TB
Full platform launch. Open source SDK released. Teams self-hosting Aiviber on Filecoin. Knowledge bases growing as agent usage compounds. Skill Repository mode driving cross-organization knowledge sharing.

Month 12: 25 TB
Ecosystem adoption across Web3 development teams. Multiple verticals: smart contract development, protocol operations, DeFi research, on-chain analytics. Each team's Company AI Brain a growing Filecoin data contributor.

Adoption, Reach, and Growth Strategies

Target audience:

Blockchain and Web3 development teams using AI agents for smart contract development, protocol operations, DeFi research, and on-chain analytics. AI infrastructure developers building agent-powered applications who need a verifiable, permanent memory layer. Protocol infrastructure operators managing multiple networks who need agents that retain operational context across sessions.

Why these teams are reachable:

Nodefleet operates across all of these verticals. We run RPC infrastructure for 42 networks, operate 33+ protocol networks, and build AI agent tooling for our own operations. The teams we work alongside daily are the primary target audience for Aiviber. We see the agent memory problem firsthand because we experience it ourselves.

Onboarding strategy:

First 10 users come from Nodefleet's existing protocol infrastructure community. Teams already using AI agents for blockchain development who have experienced the stateless agent problem directly. The MCP integration means adoption requires no workflow change developers connect Aiviber to tools they already use.

First 100 users come from the open source SDK release and direct outreach to Web3 developer communities. Any team can self-host Aiviber on Filecoin using the SDK, making self-service adoption possible without a relationship with Nodefleet.

Growth beyond 100 users is driven by the compounding nature of the product: teams that use Aiviber longer get more value from it, which increases retention and word-of-mouth. Teams that share Skill Repository entries with other teams create network effects that make the platform more valuable with each new member.

Development Roadmap

Milestone 1: Company AI Brain + Filecoin Deal Engine + MCP Integration (Weeks 1–5)

Functionality delivered:
Core Company AI Brain with Filecoin-backed storage. Automated deal-making for all knowledge writes. MCP integration with Claude, Cursor, and VS Code. Session Memory and Project Memory modes live. Internal testing with Nodefleet's own agent operations.

Specific deliverables:
Company AI Brain storage and retrieval engine with Filecoin deal-making
MCP server implementation for Claude, Cursor, VS Code, and Windsurf
Session Memory mode with configurable TTL and automatic expiry
Project Memory mode scoped to specific repositories or protocols
Filecoin deal engine: automated provider selection, CID generation, deal tracking
Minimum 3 storage providers per knowledge entry
Internal testing with live agent sessions from Nodefleet operations

Team: 2 backend engineers (memory engine, Filecoin integration), 1 engineer (MCP server implementation)

Timeline: August 1 to September 5, 2026

Funding: $20,000

Milestone 2: Team Knowledge Base + Skill Repository + Memory Dashboard (Weeks 6–10)

Functionality delivered:
Team Knowledge Base for persistent cross-project intelligence. Skill Repository for versioned agent capabilities. Memory Dashboard with Filecoin deal visibility. Hot cache retrieval layer for performance.

Specific deliverables:
Team Knowledge Base: persistent storage shared across all agents and teammates, backed by Filecoin
Skill Repository: versioned agent capabilities with sharing across the organization
Hot cache retrieval layer with under 100ms target latency for frequently accessed knowledge
Memory Dashboard: Company AI Brain contents, Filecoin deal status, CIDs, verification, retrieval analytics
Cross-session continuity: agent picks up exactly where the previous session ended
Knowledge search and structured retrieval API

Team: 2 backend engineers (knowledge base, skill repository, retrieval layer), 1 frontend engineer (Memory Dashboard)

Timeline: September 8 to October 10, 2026

Funding: $18,000

Milestone 3: Open Source SDK + Public Launch + First 10 External Teams (Weeks 11–14)

Functionality delivered:
Open source Aiviber SDK under MIT/Apache-2. Public launch with documentation and self-service onboarding. First 10 external teams live with Filecoin-backed agent memory.

Specific deliverables:
Open source Aiviber SDK published to GitHub under MIT/Apache-2
Complete documentation: setup guide, MCP integration reference, Filecoin storage configuration, self-hosting guide
Self-service onboarding with free tier for first team deployment
10 external teams live with active Company AI Brains on Filecoin
Public metrics dashboard showing total Aiviber knowledge on Filecoin network
Blog post and technical writeup on Filecoin as AI agent memory infrastructure

Team: 2 engineers (SDK preparation, documentation, onboarding flow), 1 DevRel (documentation, community)

Timeline: October 13 to November 7, 2026

Funding: $12,000

Total Budget Requested

Maintenance and Upgrade Plans

Aiviber is infrastructure Nodefleet uses for its own operations. The team that builds it continues operating it. Maintenance is self-motivated.

Ongoing maintenance includes Filecoin deal renewal before expiry, hot cache performance monitoring, MCP compatibility updates as Claude, Cursor, and VS Code release new versions, security patching, and response to issues from active users.

Planned upgrades after the grant period:

FVM integration: smart contract-based knowledge verification, allowing teams to prove on-chain what their agents knew and when.

Cross-team knowledge marketplace: teams can publish and subscribe to Skill Repository entries from other organizations, creating a decentralized marketplace for agent capabilities backed by Filecoin.

On-chain context enrichment: automatic enrichment of agent memory with on-chain data, so agents that work with specific protocols have their knowledge automatically updated as the protocol evolves.

Retrieval optimization: improved semantic search across the Company AI Brain using embeddings stored on Filecoin.

Team

Team Members

• Kael Abbott, CEO


• Lowell Abbott, Founder


• Carlos Pena, COO


• Jonatan Maria, CTO

Team Member LinkedIn Profiles

• https://www.linkedin.com/in/kael-abbott-ab578713a


• https://www.linkedin.com/in/lowell-abbott-vidal-8280aa72


• https://www.linkedin.com/in/carlos-peña-valera-9b592aa3


• https://www.linkedin.com/in/ing-jonatan-fco-mar%C3%ADa-castro-179aa47b

Team Website

https://nodefleet.org

Relevant Experience

Nodefleet has been building and operating blockchain infrastructure since December 2021. We operate across RPC infrastructure, protocol operations, AI agent tooling, and staking. We serve 18B+ monthly requests across 42 networks and run 33+ protocol networks in production.

Aiviber is not a new concept for us. We built and operate an agent memory system for Nodefleet's own development and infrastructure teams. The grant funds the productization of that system with Filecoin as the storage backend, and its release as an open source platform for the broader ecosystem.

Our team has direct experience with MCP integration, having built MCP servers for internal tooling. We have operated decentralized storage infrastructure, informing the Filecoin deal engine design. And we run the AI agent workflows that Aiviber is designed to serve, which means every design decision comes from operational experience rather than product research.

Team Code Repositories

https://github.com/nodefleet

Additional Information

How we learned about the Open Grants Program:
Through active participation in the Filecoin and decentralized storage ecosystem. As infrastructure operators across multiple blockchain verticals, we follow Protocol Labs and Filecoin Foundation programs closely.

Best contact email: nodes@nodefleet.net

Additional context:

Aiviber addresses a problem that every team using AI agents experiences but that no one has solved on decentralized storage. The grant funds a specific technical contribution: making Filecoin the verifiable, permanent storage layer for AI agent intelligence.

The open source SDK released at Milestone 3 is a genuine public good. Any team building AI agent applications can use it to store agent memory on Filecoin without depending on Nodefleet. We expect the SDK to become one of the most practical entry points for AI-focused developers into the Filecoin ecosystem.

We believe AI agent memory is one of the most important new categories of data that will be generated over the next five years. Establishing Filecoin as the storage layer for that data now, while the market is forming, is the right time to build this.

Open Grant Proposal: Aiviber Persistent AI Agent Memory on Filecoin

The proposal points to the external nodefleet/aiviber repository but names no files, tests, or entry points; start by reviewing that repository and its implementation plan. Done would require the production-ready Filecoin-backed memory platform, MCP integrations, four memory modes, dashboard, deal engine, and SDK described here.

Written by the indexing model from the issue text.