Mirrored from filpgf.io — ProPGF Batch 3 (Karma program 1479, application 6a1892d05513a8d9decf00d5, status: pending). Contact details redacted; canonical application lives on filpgf.io. 1.1 Project Name Fetcher — Filecoin Agent SDK 1.2 Project Github https://github.com/Klorenn/LavanderSDK 1.3 Project Website https://lavander-sdk.vercel.app 1.4 Team Lead/Point of Contact Pau Andreu Koh Cuende 1.5 Category [ "Core Infrastructure" ] 1.6 Open Source Status Fully Open Source 2.1 Project Summary Fetcher is a TypeScript SDK that gives AI agents — Claude, GPT-4o, Gemini — 17 native tools to store, retrieve, verify, and remember data on Filecoin Onchain Cloud. It is the first SDK to bridge MCP, LangChain, and LlamaIndex with Synapse SDK in a single package. The project fills a critical ecosystem gap: no existing tool gives AI agents structured, persistent, cryptographically-verifiable storage on Filecoin with zero blockchain knowledge required from the developer. Fetcher has been built over 55+ commits, spans 7 packages, passes 75 tests, and is already deployed as a live documentation site. The SDK is ready for production use on Calibration testnet and designed for mainnet with safety defaults that block all paid operations until explicitly enabled. 2.2 Who does this work support? [ "Application Users", "Application Builders", "Storage Providers", "Network Infrastructure" ] 2.3 Total Funding Requested (USD) 100000 2.4 Milestones & Budget [ { "title": "Core SDK, MCP Server, and Agent Memory System", "description": "Deliverable: The full Fetcher SDK core published to npm with 11 tools across 2 adapters. MCP stdio server exposing storage, verification, and memory tools. Agent memory system with versioned, TTL-aware persistence between sessions. Unit tests passing with fake backends (no wallet required).", "dueDate": "2026-08-31", "fundingRequested": "50000", "completionCriteria": "- `@fetcher-fil/core` and `@fetcher-fil/mcp` published to npm at v0.1.0\n- 11 tools functional via MCP: store_file, retrieve_file, list_files, delete_file, verify_cid, check_deal, get_proof, store_memory, retrieve_memory, update_memory, list_memories\n- Agent memory system with version tracking, TTL expiration, and fallback values\n- Spending policy enforced: paid operations blocked by default, confirmation required\n- All tests passing with FakeStorageBackend and MemoryIndexBackend\n- Demo video: agent storing data and remembering context between sessions" }, { "title": "Multi-Framework Support, Observability Tools, and Public Launch", "description": "Deliverable: Fetcher available on all 4 frameworks (MCP, LangChain, LlamaIndex, SDK). Observability tools for cost estimation, balance checking, and storage stats. Full documentation site deployed. v1.0.0 published to npm. Community announcement.", "dueDate": "2026-08-31", "fundingRequested": "50000", "completionCriteria": "- `@fetcher-fil/langchain`, `@fetcher-fil/llamaindex`, and `@fetcher-fil/sdk` published to npm at v1.0.0\n- All 17 tools available identically across all 4 adapters\n- 4 observability tools: get_balance, estimate_cost, get_storage_stats, list_deals\n- Fetcher SDK class with fluent API and memory namespace\n- Documentation site live at lavander-sdk.vercel.app with 9 sections\n- Minimum 5 verified paid storage deals completed on Calibration testnet\n- All 75+ tests passing across 7 packages\n- Community announcement posted in Filecoin Slack (#fil-dev) and Filecoin Forum\n- Final ProPGF report submitted with on-chain verification links" } ] 3.1 Impact pathway Fetcher → AI developers install SDK (output) → AI agents store data on Filecoin instead of S3/IPFS (outcome) → More paid deals, more USDFC locked, more storage provider revenue (impact on Objective 2 KPIs). The pathway is: lower barrier → more developers → more agents → more deals → stronger network economics. 3.2 Verification metrics | Metric | Data source | How measured | Target | |---|---|---|---| | Monthly active @fetcher-fil npm downloads | npm registry | npm API | 500+ monthly | | GitHub stars | GitHub API | Repository stars | 100+ | | Deals created via SDK | Synapse/Filecoin chain | Tagged transactions | 500+ | | Community integrations | GitHub search | Repos importing fetcher-fil | 50+ | 3.3 References **Filecoin Slack community** — Developers who have expressed interest in MCP + Filecoin integration 2. **Synapse SDK team** — `@filoz/synapse-sdk` is the core dependency; Fetcher drives adoption of their SDK 3. **AI agent developers** — LangChain and LlamaIndex community members seeking Filecoin storage solutions 4.1 Monthly Operating Burn [ "< $10K (basic solo operation or part-time team)" ] 4.2 What % of total team monthly burn depends on this grant? 100% 4.3 If this grant is not awarded, what happens? The SDK is already built and functional as open-source software. Without funding, development continues at a significantly slower pace — nights and weekends only. The npm publication, documentation site maintenance, community support, and the planned Python SDK are deprioritized indefinitely. The core TypeScript SDK remains available on GitHub under MIT license for anyone to use or fork, but without dedicated time for community growth, onboarding support, or responding to Synapse SDK updates. The risk is that the AI agent ecosystem moves fast — LangChain and MCP are growing exponentially — and without dedicated resources, Filecoin misses the window to establish itself as the default storage backend for AI agents before centralized alternatives capture the market. 4.4 Core Team **Kl0ren** — Lead Developer & Sole Contributor - Built the entire Fetcher SDK: 55+ commits, 7 packages, 75 tests, full documentation site - Full-stack TypeScript: React, Vite, Tailwind, Framer Motion, Node.js - Synapse SDK integration, MCP protocol implementation, LangChain/LlamaIndex tool development - Filecoin Calibration testnet validation with real on-chain transactions - Open-source maintainer with published projects on GitHub 4.5 Has your team received a ProPGF grant or funding from PLFIF before? [ "No" ] 5.1 Key risks & dependencies | Risk | Mitigation | | Synapse SDK breaking changes | Pinned dependency (`^0.41`). Fetcher's StorageBackend interface isolates Synapse — only one file needs updating if the API changes. | | Calibration testnet instability | SDK defaults to Calibration but is designed for mainnet. Error handling surfaces chain issues clearly to the agent. | | Low MCP adoption for Filecoin | Fetcher supports 4 frameworks (MCP, LangChain, LlamaIndex, SDK) — not dependent on any single one. | | Solo developer bus factor | Code is open-source (MIT). Architecture is documented. Community can fork and maintain. | Anything else you want to share that we didn't ask? Fetcher has been built entirely as a solo open-source project over the past weeks. The codebase is clean, tested, documented, and deployed. The SDK already works against Calibration testnet with real Synapse transactions. What remains is: publishing to npm, growing community adoption, adding Python SDK support, and building the payment rails abstraction. ProPGF funding would accelerate all of these and ensure Filecoin has a first-class SDK in the AI agent ecosystem before centralized alternatives capture this market. Contributing to Core Infrastructure? Fetcher is the first unified SDK connecting AI agent frameworks — MCP, LangChain, LlamaIndex — to Filecoin Onchain Cloud via Synapse SDK. It provides 17 tools for storage, verification, observability, and agent memory on Filecoin. Without Fetcher, AI developers must build their own Synapse integration from scratch — a barrier that blocks thousands of AI agents from using Filecoin as their storage backend. The project depends on `@filoz/synapse-sdk` and serves any AI developer using MCP-compatible clients (Claude Desktop, Cursor, Continue, Cody), LangChain agents, or LlamaIndex agents. Objective 1 Direct Objective 2 Indirect Objective 3 Direct Open Source Context Not applicable — Fetcher is fully open source under the MIT License. All 7 packages (core, mcp, langchain, llamaindex, sdk, testkit, landing) are publicly available on GitHub with no closed components or proprietary dependencies.