SkillAgentSearch skills...

omni-mcp-gateway

Rust Enterprise MCP Fleet, Tool Groups & Code Mode Gateway v6.0 (macOS, Linux, Windows x64/ARM64): rustls TLS/mTLS, OIDC/JWKS (RS256/ES256/EdDSA), tokio async, SQLite WAL+CAS, wasmi Wasm sandbox, Okapi BM25 0-turn routing, SSA parallelization, Saga rollbacks & Fleet UI.

Install / Use

claude mcp add Rangow4562 -- npx -y github:Rangow4562/omni-mcp-gateway

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

74/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of omni-mcp-gateway

omni-mcp-gateway scores 74/100 on our quality scale, 853rd of 957 AI & Machine Learning skills we index.

Its MCP Server is 40 KB long, well organised into 47 sections with 13 code examples: long enough that it reads more like full documentation than a focused instruction file, which agents can find harder to follow.

It has 3 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
21/30
Structure
20/20
Description
15/15
Adoption
3/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated yesterday, so omni-mcp-gateway is actively maintained.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 80/100, with 2 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-11-16. Automated pattern scan on 2026-11-16. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

omni-mcp-gateway compared with similar skills

All 4 of these similar skills score higher than omni-mcp-gateway; compare them before choosing.

SkillScoreStarsUpdatedFormat
omni-mcp-gateway (this skill)by Rangow45627431d agoMCP Server
claude-memby thedotmack100117.7ktodayCLAUDE.md
Agent-Reachby Panniantong100112.4k9d agoCLAUDE.md
Understand-Anythingby Egonex-AI10092.9k1d agoCLAUDE.md
rufloby ruvnet10081.9ktodayCLAUDE.md

Frequently asked questions

How do I install omni-mcp-gateway?
Run claude mcp add Rangow4562 -- npx -y github:Rangow4562/omni-mcp-gateway. The install tabs above show the steps for each supported agent.
Which AI agents does omni-mcp-gateway work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is omni-mcp-gateway safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 80/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is omni-mcp-gateway still maintained?
The repository was last updated yesterday, so omni-mcp-gateway is actively maintained.
<p align="center"> <img src="assets/omni-mcp-eink-banner.png" alt="Omni MCP Banner" width="100%" /> </p>

Omni MCP (omni-mcp v5.5 — Autonomous Runtime & Enterprise MCP Gateway)

npm version npm downloads License CI/CD M8ven Score

A high-performance, zero-dependency Rust enterprise Model Context Protocol (MCP) management platform and runtime gateway combining native in-process Omni Code Mode with enterprise service mesh governance, Okapi BM25 intent ranking, columnar predicate pushdown, resumable continuation snapshots, and declarative Saga compensations.

Omni MCP unifies hundreds of microservices and thousands of tools behind a standardized, sub-millisecond, token-optimized Rust gateway with:

  • Native Code Mode Execution (codemode): In-process sandbox executing JavaScript/TypeScript syntax scripts with closure syntax (|x| expr) and service host calls in $\sim 0.10\text{ ms}$ ($100\ \mu\text{s}$).
  • Resumable AST Continuations: Stateless suspend/resume of execution state (SuspendedExecutionFrame) across approval gates (is_mutation: true) via durable FNV-1a checksummed WAL snapshots.
  • Zero-Copy Columnar Predicate Pushdown: Compiles .filter() and .map() AST closures directly into vectorized PhysicalExpr pipelines over contiguous columnar batches (ColumnarRecordBatch).
  • Static AST SSA Auto-Parallelization: Automatic dependency analysis lifts independent sequential await codemode.* calls into concurrent Stmt::ParallelWave batches without requiring complex LLM Promise.all emissions.
  • Declarative Saga Compensation Rollbacks: Ephemeral execution transaction logs unwind completed mutations in reverse (LIFO) order when a downstream step fails.
  • Speculative 0-Turn Schema Pre-Injection: True Okapi BM25 ($k_1 = 1.2, b = 0.75$) + stem routing injects the top-$K$ highest-confidence .d.ts declarations on Turn 1 based on natural-language user prompts.
  • Deterministic Simulation Testing (dst) Engine: 64-bit Xoshiro256StarStar seeded PRNG fault-injection suite validating agent scripts against network drops, circuit trips, latency spikes, and malformed payloads.
  • Authoritative -320xx Error Contract & Multi-Revision MCP Protocol Negotiation: Compiler-checked FeatureStatus enum mapping every runtime guardrail (-32030..-32038) and negotiating official MCP revisions (2024-11-05, 2025-03-26, 2025-06-18, 2025-11-25), with graceful fallback for future spec revisions like 2026-07-28 and stateless server/discover over Streamable HTTP.
  • Live HTTP/1.1 Downstream Transport (transport-http): Feature-gated, pure standard library connection-pooled HTTP client with timeouts, keep-alive, and jittered exponential retries feeding per-service circuit breakers and Saga rollbacks.
  • Schema Preflight Validator (-32035 SCHEMA_PREFLIGHT_FAILED): Static AST input schema type-checker detecting all violations in a single pass before execution or approval gates for single-turn LLM self-correction.
  • HMAC-SHA256 Authenticated Continuations & Approval Hardening: Tamper-proof continuation frames signed via HMAC-SHA256 with atomic compare-and-swap (CAS) single-use binding (-32038 APPROVAL_INVALID), approver identity and argument digest verification, and crash-safe WAL/snapshot Saga compensation ledger.
  • Comprehensive Sandbox Resource Limits: Strict execution guardrails bounding instruction fuel (-32032), string/array heap growth (-32036 MEMORY_EXHAUSTED), wall-clock time (-32037 TIMEOUT), and internal panic containment (catch_unwind).
  • Dynamic TypeScript Declarations (declare namespace codemode): Automatically converts OpenAPI 3.1, Protobuf, and Thrift JSON Schemas into compact .d.ts declarations, compressing tool prompt token footprints by $65%$ to $95%$.
  • Search & Execute Pattern: Built-in codemode.search(query) and codemode.describe(server, tool?) allowing agents to discover and inspect typed definitions on demand without context bloat.
  • Zero-Outbound Isolation & Approval Gates: globalOutbound: null isolation blocks raw fetch() or filesystem calls, while human-in-the-loop approval gates intercept state-mutating operations (is_mutation: true).
  • Full Mesh Governance: Wire envelope framing & protocol translation (gRPC, HTTP, simulated RPC), circuit breaking, semantic caching, ABAC security, OpenTelemetry W3C distributed tracing, canary weighted routing, and SHA-256 Merkle audit trail.


🛡️ What's New in Version 5.0 ("Trust Boundary Reset & MCP 2026-07-28 Conformance")

  • 🤝 Full MCP 2026-07-28 Protocol Conformance (B1–B10):
    • Required "resultType": "complete" on all results and "resultType": "input_required" with requestState for multi-round-trip requests (B1).
    • Deterministic lexicographical ordering and caching metadata (ttlMs: 300000, cacheScope: "workspace") across tools/list, resources/list, prompts/list, and resources/templates/list (B2).
    • Streamable HTTP protocol header validation on POST /omni/mcp and /<service>/mcp: requires Mcp-Method and Mcp-Name, rejecting mismatches with HTTP 400 and -32020 HEADER_MISMATCH (B3).
    • Pruned deprecated endpoints (ping, logging/setLevel, roots/list_changed) and validated client capabilities with -32021 MISSING_REQUIRED_CLIENT_CAPABILITY (B4).
    • Spec-reserved error code allocation (-32020..-32099 for spec, -32000..-32019 for gateway) with backward-compatible data.legacy_code mappings (B5).
    • RFC 9728 protected resource metadata at /.well-known/oauth-protected-resource, HTTP 401 WWW-Authenticate: Bearer resource_metadata=..., and HTTP 403 insufficient_scope (B6).
    • MCP Tasks extension (io.modelcontextprotocol/tasks) exposing tasks/get and tasks/update for suspended execution frame lifecycle management (B7).
    • Deprecated legacy /sse endpoint returning 410 Gone by default (opt-in --features legacy-sse) (B8).
  • 🛡️ Universal Route Authorization & Atomic Approvals (A1–A6):
    • Deny-by-default role-based access control (RBAC) across all /api/* and /api/v1/* endpoints (Agent, Operator, Approver, Admin).
    • Atomic compare-and-swap (CAS) frame consumption under write lock preventing race conditions on multi-approver resumes.
    • Separation of duties cryptographically forbidding mutation requesters from approving their own actions.
    • Length-prefixed HMAC-SHA256 continuation signing over all 12 frame fields with argument digest verification.
  • 🌐 HTTP Front-Door Hardening & DoS Containment (A7–A10, A14):
    • Recursive JSON parser depth guard (MAX_JSON_DEPTH = 64) preventing stack-overflow crashes.
    • Bounded worker channel shedding saturated load with 503 Service Unavailable, plus connection-level catch_unwind preventing thread death.
    • Persistent BufReader framing across HTTP/1.1 keep-alive loops with strict digit-only Content-Length checks.
  • ⚡ Architecture, Observability & Tooling (D7, D8, C1):
    • $O(1)$ ring-buffer trace eviction via VecDeque<TraceSpan>.
    • Versioned /api/v1/* REST routes and OpenAPI 3.1 specification at /openapi.json.
    • CLI subcommand direct execution (omni-gateway list, omni migrate).

🤝 What's New in Version 4.7 ("Protocol Parity & Real Wire")

  • 🤝 MCP Protocol Parity (2024-11-05, 2025-03-26, 2025-06-18, 2025-11-25) (M1): First-class protocol capability matrix supporting four active revisions with graceful fallback for 2026-07-28. Implements stateless server/discover, subscriptions/listen, _meta protocol version and client capability negotiation, batch gating, W3C trace context bridging, and Streamable HTTP on /omni/mcp and /<service>/mcp with deprecated legacy /sse.
  • 🌐 Live HTTP/1.1 Downstream Transport (transport-http) (M2): Feature-gated, pure standard library HTTP/1.1 downstream client with connection pooling, keep-alive, connect/read/write timeouts, and bounded retries with jitter feeding the per-service circuit breaker and Saga unwinder.
  • 📋 Schema Preflight Validator (-32035 SCHEMA_PREFLIGHT_FAILED) (M3): Static AST input schema type-checker validating parameters against registered schemas before approval gates or execution, collecting all violations in a single pass for single-turn LLM correction.
  • 🔐 Approval & Continuation Hardening + Persistent Saga Ledger (M4): Authenticated continuation frames via HMAC-SHA256, atomic compare-and-swap single-use approval binding (-32038 APPROVAL_INVALID), OS entropy mixing, and crash-safe WAL/snapshot persistence of in-flight Saga compensations.
  • 🧱 Sandbox Resource Limits (M5): Comprehensive sandbox boundaries with memory limit tracking (-32036 MEMORY_EXHAUSTED), wall-clock execution timeouts (-32037 TIMEOUT), and panic containment (catch_unwind).
  • 🛡️ HTTP Surface Hardening & Universal RBAC (H1): DNS rebinding protection via Origin validation, loopback bind by default with non-loopback startup guard, universal Bearer authentication and role-based access control (RBAC) across all /api/* endpoints, slowloris timeouts, header line and size caps, and HTTP request smuggling protection.

🛡️ What's New in Version 4.6 ("Honesty & Hardening")

  • 🧪 Code Mode Test Fortress (M1): In-module #[cfg(test)] unit test suites across codemode.rs, json.rs, wal_store.rs, and gateway.rs verifying parser diagnostics, evaluator closures, fuel exhaustion (-32032), globalOutbound: null isolation (-32033), vectorized PhysicalExpr evaluation, SuspendedExecutionFrame round-trip resume identity, FNV-1a WAL corruption rejection, and Saga LIFO unwinding order.
  • 🎯 True Okapi BM25 Ranking (M2): Replaces additive substring scoring with full Okapi BM25 ($\text{IDF}$ corpus weighting, term-frequency saturation $k_1 = 1.2$, document length normalization $b = 0.75$, camelCase/snake_case splitting, suffix stemming, and field boosts) across speculative_pre_inject, discover_server, and recommend_tools.
  • 📐 Authoritative -320xx Error-Code Contract (M3): Strongly-typed FeatureStatus enum and structured RPCError contract covering -32030 CIRCUIT_OPEN, -32031 APPROVAL_REQUIRED, -32032 FUEL_EXHAUSTED, -32033 OUTBOUND_BLOCKED, -32034 SAGA_COMPENSATED, and -32035 SCHEMA_PREFLIGHT_FAILED.
  • 🤝 MCP Protocol Version Negotiation (M4): Lifecycle negotiation on initialize echoing client protocolVersion when in {"2024-11-05", "2025-03-26"} (defaulting to "2025-03-26"), plus additive 2025-03-26 tool capability annotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint).
  • 🧹 Single-Sourced Versioning & Documentation Honesty: Rust runtime versions single-sourced from env!("CARGO_PKG_VERSION") with an automated drift-guard test across Cargo.toml, package.json, ui/package.json, and README.md.

🚀 What's New in Version 4.5

  • 🔄 Resumable AST Continuations for Approval Gates (Stateless Suspend/Resume): Serializes instruction pointer, scope heap, and Merkle hashes into Crockford Base32 ULID (SuspendedExecutionFrame) in durable WAL when hitting APPROVAL_REQUIRED, resuming at instruction_ptr + 1 upon approval without re-running read tools.
  • ⚡ **Zero-Copy Columnar

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars3
CategoryAI
Updated1d ago
Forks2

Languages

Rust

Trust signals

80/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 medium1 low