Ultra-lean, zero-third-party-dependency, production-grade Model Context Protocol (MCP) server implementations directly on raw JSON-RPC 2.0 over standard I/O (
stdio).
Official MCP SDKs bundle substantial runtime dependencies (Pydantic, AnyIO, Starlette, websockets) that introduce supply chain vulnerability surface, binary bloat, and deployment friction in constrained or enterprise environments.
ZeroMCP strips out all middleware bloat, implementing the pure wire protocol directly with:
- Zero Third-Party Packages: Pure standard library or minimal native async primitives.
- Sub-Millisecond Execution: 10 to 40 microsecond tool dispatch latency (~35,000–50,000 calls/sec).
- Strict Transport Isolation: Standard output (
stdout) reserved exclusively for JSON-RPC 2.0 framing; diagnostics routed strictly tostderr. - Scoped Security & Delegation: Sub-5 microsecond ephemeral capability tokens (
zeromcp_token.py) with hierarchical multi-tenant scopes and negation (["tools:*", "!tools:execute_shell"]). - Rigorous Verification: 43 automated wire conformance checks, 7 FGAC security tests, and 24 supervisor regression tests asserting full spec compliance.
| Implementation | 3rd-Party Deps | Handshake Latency (Cold) | Per-Call Latency (Mean) | Throughput (Single-Core) | Baseline Memory (RSS) |
|---|---|---|---|---|---|
ZeroMCP Python (zeromcp.py) |
0 | ~26 ms | ~24 µs | ~36,000+ calls/sec | ~14 MB |
ZeroMCP Rust Native (zeromcp) |
0 (runtime) | < 2 ms | < 6 µs | ~120,000+ calls/sec | ~4 MB |
Official Python SDK (mcp + Pydantic) |
34+ packages | ~140 ms | ~450 µs | ~2,200 calls/sec | ~48 MB |
(Run python3 benchmarks/benchmark.py to reproduce locally).
Add ZeroMCP directly to your Claude Desktop configuration:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"zeromcp": {
"command": "python3",
"args": ["/ABSOLUTE/PATH/TO/ZeroMCP/zeromcp-py/zeromcp.py"]
}
}
}Add to your project workspace root at .cursor/mcp.json:
{
"mcpServers": {
"zeromcp": {
"type": "stdio",
"command": "python3",
"args": ["/ABSOLUTE/PATH/TO/ZeroMCP/zeromcp-py/zeromcp.py"]
}
}
}ZeroMCP eliminates awkward JSON piping with built-in developer CLI tools:
# List all registered tools & schemas in an aligned table
python3 zeromcp-py/zeromcp.py --list
# Call tool directly with human shorthand (no JSON escaping required)
python3 zeromcp-py/zeromcp.py --call eval_math "42 * 1024 / 8"
# Launch interactive REPL console for ad-hoc tool testing
python3 zeromcp-py/zeromcp.py --repl
# Print auto-resolved claude_desktop_config.json block
python3 zeromcp-py/zeromcp.py --print-config
# View CAS storage metrics or prune old cached blobs
python3 zeromcp-py/zeromcp.py --cas-stats
python3 zeromcp-py/zeromcp.py --cas-gc 7Keep agent context windows lean without third-party sidecars:
-
Content-Addressed Storage (CAS): Tool responses
$\ge 4\text{KB}$ are automatically saved to local sharded storage (~/.zeromcp/cas/or local fallback). ZeroMCP returns a compactcas://sha256/<hash>token pointer + preview, reducing context consumption by up to 95%. Agents can inspect slices using built-incas_fetchandcas_searchtools. -
Command Output Distiller: Pre-filters verbose test runners (
pytest,cargo test,git status) before feeding stdout to the model:python3 zeromcp-py/command_distiller.py --cmd "pytest" --telemetry # 📊 [Distiller: Saved 64.2% tokens (12,400 -> 4,440 chars)]
Verify raw wire compliance in one line:
printf '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1"}}}\n{"jsonrpc":"2.0","method":"notifications/initialized"}\n{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"echo","arguments":{"message":"hello world"}}}\n' | python3 zeromcp-py/zeromcp.pyZeroMCP is a cornerstone protocol primitive in the CreativeUnCommons agent architecture suite. Each tool in the ecosystem is engineered under a strict Zero-Third-Party-Dependency Mandate (Python standard library only), Sub-Millisecond Execution Budgets, and Deterministic Correctness. They can be deployed independently or composed into a hardened, enterprise-grade autonomous runtime:
flowchart TD
subgraph S1["1. Durability & State Machine"]
ZL["⚡ <b>ZeroLoop</b><br/>• Event-Sourced SQLite WAL<br/>• Idempotent Tool Execution<br/>• Deterministic Crash Recovery"]
end
subgraph S2["2. Privacy & Compliance Boundary"]
ZLeak["🛡️ <b>ZeroLeak</b><br/>• Sub-50µs Reversible PII Masking<br/>• Ephemeral In-Memory Vault<br/>• Prompt-Injection Firewall"]
end
subgraph S3["3. Context & Cost Routing"]
HG["🚪 <b>HyperGate</b><br/>• In-Memory BM25 Tool Pruning (<1ms)<br/>• System-1 Fast Path (90/10 Split)<br/>• Budget Ceiling Circuit-Breaker"]
end
subgraph S4["4. Tool Execution & Wire Protocol"]
ZMCP["🔌 <b>ZeroMCP</b> (This Package)<br/>• Raw JSON-RPC 2.0 Stdio/SSE Core<br/>• Content-Addressed Storage (CAS)<br/>• Scoped Capability Delegation"]
end
subgraph S5["5. Universal Observability"]
AT["🌲 <b>AgentTrace</b><br/>• W3C Distributed Context (traceparent)<br/>• OpenTelemetry GenAI Semantics<br/>• Sub-ms JSONL / SQLite & OTLP Exporter"]
end
Task["User / Autonomous Task"] --> ZL
ZL -->|"Step 1: Ingest State"| ZLeak
ZLeak -->|"Step 2: Redacted Safe Prompt"| HG
HG -->|"Step 3: Pruned Tool Schemas"| Model["Cloud / Local LLM"]
Model -->|"Step 4: Tool Call"| ZMCP
ZMCP -->|"Step 5: CAS Output Pointer"| ZLeak
ZLeak -->|"Step 6: Restored Data"| ZL
ZL -->|"Step 7: Verified Result"| Done["Task Completion"]
ZL -.->|"State Transitions"| AT
ZLeak -.->|"Sanitization Spans"| AT
HG -.->|"Token & Cost Metrics"| AT
ZMCP -.->|"Tool IPC Spans"| AT
| Package | Role in Unified Stack | Standalone Capability | What You Achieve Together |
|---|---|---|---|
| ⚡ ZeroLoop | Durable Orchestrator | Event-sourced state machine with crash recovery & async human pauses. | Tasks resume after crashes in |
| 🛡️ ZeroLeak | Privacy Sentinel | Sub-$50\text{µs}$ reversible PII & credential redaction with in-memory vaulted restitution. | Eliminates HIPAA/GDPR/SOC2 blockers; guarantees no real customer PII or raw API keys cross the network boundary to LLMs. |
| 🚪 HyperGate | Inference Router & Tool Pruner | Sub-millisecond BM25 inverted index tool pruner, System-1 classifier, and budget kill-switch. | Shrinks 500+ tool schemas down to relevant sets in |
| 🔌 ZeroMCP | Tool Protocol & CAS Engine (This Package) | Zero-dependency Model Context Protocol (MCP) runtime (stdio/SSE) with CAS storage. | High-speed ( |
| 🌲 AgentTrace | Distributed Telemetry & Auditing | W3C traceparent context propagation, GenAI semantic conventions, and native OTLP export. | Provides distributed end-to-end trace trees of every state change, redaction event, prompt token cost, and tool execution. |
| Implementation | Distribution Target | Runtime Dependencies | Conformance Status | User Compilation Required? |
|---|---|---|---|---|
zeromcp-py |
Standalone script (zeromcp.py) |
Pure Python stdlib (sys, os, json, ast, math) |
✅ 43/43 (100%) | No (Direct python3 zeromcp.py) |
zeromcp-ts |
Standalone script (zeromcp.ts) |
Pure ECMAScript (Node 22+, Bun, Deno) | ✅ Verified (CI Node 24) | No (Direct bun or node --experimental-strip-types) |
zeromcp-rs |
Pre-compiled static binaries (dist/) |
Zero (self-contained native binary) | ⚙️ 6-Target CI Release Matrix | No (Pre-built via GitHub Actions) |
zeromcp.wasm |
Universal WASI module | wasmtime, wasmer, or Node WASI runner |
⚙️ WASI Preview 1 Bytecode | No (Pre-built universal bytecode) |
For comprehensive client setup (Claude Desktop, Cursor, VS Code, Docker, WebAssembly), consult the ZeroMCP Complete Usage & Integration Guide. For release history and security updates, see the Changelog.
Run the cross-language compliance suite (43 checks):
python3 tests/conformance_test_runner.pyRun the stdio rate-limiting supervisor test suite (24 regression tests):
python3 tests/test_stdio_limiter.pyValidates:
- Protocol Handshake (
initialize) & Capability Negotiation - Notifications (
notifications/initialized- verifying zero response leakage) - Diagnostic Heartbeat (
ping) - Tool Discovery (
tools/list) & Schema Validation - Tool Execution (
echo,eval_math,system_info) - Error Taxonomy (
-32700,-32600,-32601,-32602,-32603) - Framing & 10 MiB Bounds
- Token-Bucket Throttling, Batch Aggregation & Clean EOF Drain