TinyZKP
@logannye
About TinyZKP
A height-compressed, sublinear-space STARK prover. Designed to make very long zero-knowledge proofs practical on fixed-memory hardware (CPUs, GPUs, & edge devices). Post-quantum. Built in Rust.
Basic information
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"tinyzkp": {
"type": "http",
"url": "https://mcp.tinyzkp.com"
}
}
}Tools
No tools detected
We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.
Overview
What is TinyZKP?
TinyZKP mints a tamper-evident zero-knowledge proof that a state-transition chain is consistent — start at X, apply these steps, provably reach Y — so an agent or API caller can hand over a verifiable receipt instead of requiring trust. It provides transparent, post-quantum STARK-style proofs with an O(√T) prover-memory architecture, no trusted setup, and verification in milliseconds. It serves developers, AI agents, and API users who need verifiable state-transition receipts.
How to use TinyZKP?
Install the MCP server for AI agents (Claude, GPT, Cursor) via claude mcp add --transport http tinyzkp https://mcp.tinyzp.com, then call its proof tools as native function calls. Alternatively, use the terminal CLI via npx @tinyzp/cli or the plain HTTP API at https://api.tinyzzp.com with an API key from tinyzzp.com/signup. Verification is always free and works in the browser via the WASM package @tinyzzp/verify.
Key features of TinyZKP
- O(√T) prover memory vs standard STARK O(T) — ≈4,096× reduction measured
- Verification in polylog time, always free
- Transparent (no trusted setup) and post-quantum (hash-based)
- SDKs for Python, TypeScript, and Rust
- MCP server for native AI agent integration (Claude, GPT, Cursor)
- Browser WASM verifier with no server or API key required
Use cases of TinyZKP
- State machine attestation: prove a sequence of state transitions without revealing intermediate values
- Audit-log checkpoints: generate a tamper-evident receipt that a log transitioned from one hash to another through a series of entries
- AI agent verifiable receipts: an agent hands over a proof that its chain-of-thought or action sequence reached a stated outcome
- API state transition proofs: two systems prove they agree on a sequence of values without exposing the data itself
FAQ from TinyZKP
What does TinyZKP prove that other schemes don't?
TinyZKP proves state-transition chain consistency with a transparent (no trusted setup) STARK that uses an O(√T) prover-memory architecture, unlike standard STARKs which require O(T) memory. This makes long traces practical without reserving high-RAM servers.
What are the runtime dependencies?
The CLI requires Node 18+ ESM. The Python SDK works with Python 3 and pip install tinyzkp. The TypeScript SDK ships a dual ESM+CJS build for Node 18+, bundlers, and edge runtimes. The WASM verifier works in any modern browser.
Where does proof data live?
Proving jobs are submitted to the hosted API at api.tinyzp.com; job and proof data reside on TinyZKP's servers during the lifecycle. Verification can be done client-side via the WASM verifier without sending proof data to any server.
What transports and authentication are available?
The MCP endpoint supports HTTP transport and is publicly rate-limited (no API key required). The full HTTP API uses HTTPS with bearer token authentication (API key tzk_...). The WASM verifier and CLI can verify proofs locally with no network call.
What are the known limits?
Free tier: 100 proofs per month, 1 in‑flight job, 10 RPM, $5/mo cap. Developer plan: 4 in‑flight, 100 RPM, $500/mo cap. Proofs under 10K trace steps cost $0.05/proof; longer traces cost more per step. Verification is always free and unlimited.
More Memory & Knowledge MCP servers
Semantic Scholar MCP Server
YUZongminA FastMCP server implementation for the Semantic Scholar API, providing comprehensive access to academic paper data, author information, and citation networks.
Notion MCP Integration
danhilseA simple MCP integration that allows Claude to read and manage a personal Notion todo list
Obsidian MCP Server
cyanheadsRead, write, search, and surgically edit Obsidian vault notes, tags, and frontmatter via MCP. STDIO or Streamable HTTP.
Solomd
zhitongblogA markdown editor — and the bridge to your LLM. Local-first, MIT, ~15 MB. Bundled MCP server lets Claude Code / Codex / Cursor drive your vault directly. 14 AI providers BYOK.
MCP Apple Notes
RafalWilinskiTalk with your notes in Claude. RAG over your Apple Notes using Model Context Protocol.
Comments