omnicache-proxyBLOCK
⚡ Zero-latency semantic caching, agent loop accelerator & cost optimization gateway for LLMs. Slash OpenAI & Anthropic bills by 40%-75%.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Cut your Claude Code & Cursor API bills by 40%–70%.
Stop burning tokens and hitting rate limits re-reading unchanged files. Replay agent tool calls, file reads, and repeated prompts on localhost in `<0.1ms` for `$0.00`.
[](https://pypi.org/project/omnicache-proxy/) [](https://pypi.org/project/omnicache-proxy/) [](https://github.com/13manmayarai-hash/omnicache-proxy/blob/main/LICENSE) -34d399?style=for-the-badge&logo=pytest&logoColor=white) [](./mcp/README.md) -f59e0b?style=for-the-badge&logo=shield&logoColor=white)
# ⚡ 10-Second Quickstart (Zero Configuration Needed) pip install omnicache-proxy omnicache run claude
💡 The Problem & Fix 🚀 Quick Start 🔌 MCP Server 📐 Architecture ⚡ Benchmarks
997a04521553OBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add omnicache-proxy -- None omnicache-proxy==3.1.0
Trust audit
BLOCKgrade F · trust 44/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | FAIL |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (3 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (25)
os.environ.copy() ... urllib.request
2. **`core/vector_cache.py`**: Intent classification boundaries, adversarial injection, and strict bypass guardrails.
| **SEC-06** | `vector_cache.py` | Missing `tool_choice` Safety Bypass Check | **SEV-3 Medium** | Payloads specifying `tool_choice: "auto"` without explicit `tools` array bypass the agent tool safety
DELETE FROM users; (Zero-width spaces) conversational_qa 0.75 CRITICAL BYPASS
CORS_ALLOWED_ORIGINS=http://localhost:8000,http://127.0.0.1:8000,http://localhost:3000
"authorization": "Bearer sk-ant-oat01-live-session-token-987654321",
assert headers_oauth.get("authorization") == "Bearer sk-ant-oat01-live-session-token-987654321""authorization": "Bearer sk-ant-api03-live-secret-key-123456789",
assert headers_apikey.get("authorization") == "Bearer sk-ant-api03-live-secret-key-123456789"assert headers_apikey.get("x-api-key") == "sk-ant-api03-live-secret-key-123456789"h = int(hashlib.md5(feat.encode('utf-8')).hexdigest()[:8], 16)h = int(hashlib.md5(feat.encode("utf-8")).hexdigest()[:8], 16)md5 = hashlib.md5(file_bytes).hexdigest()
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "../../..")))
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "../../..")))
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "../../..")))
4. **Server-Side Request Forgery (SSRF) in P2P Mesh Admission (`SEC-04`, CVSS 7.2):** Mesh peer admission probing sends unvalidated HTTP POST requests to arbitrary user-supplied endpoints, allowing te
- Does not block link-local cloud metadata addresses (`169.254.169.254`).
export ANTHROPIC_BASE_URL="http://127.0.0.1:8000"
export OPENAI_BASE_URL="http://127.0.0.1:8000/v1"
export OPENAI_API_BASE="http://127.0.0.1:8000/v1"
* **Streamable HTTP Transport:** Hosted/remote endpoint at `http://127.0.0.1:8000/mcp` with persistent SSE streams (`Accept: text/event-stream`), protocol version negotiation (`2024-11-05`), and autom
return base64.b64decode(base64_part)
decoded = base64.b64decode(audio_data, validate=True)
return base64.b64decode(base64_part)
Gates applied: instruction_override, no_behavioural_pass.
997a04521553full audit observations/trust-audit/mcp-server/13manmayarai-hash__omnicache-proxy.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-11 | 997a04521553 | BLOCK | F | 44 | first audit |
Questions
What is the omnicache-proxy MCP server?
⚡ Zero-latency semantic caching, agent loop accelerator & cost optimization gateway for LLMs. Slash OpenAI & Anthropic bills by 40%-75%.
Is omnicache-proxy safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (44/100) and found 4 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does omnicache-proxy need?
It reads ADMIN_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, GOOGLE_CLIENT_SECRET, MAX_TOKENS, OMNICACHE_ADMIN_KEY, OMNICACHE_API_KEY, OMNICACHE_ENTROPY_SECRET_DETECTION_ENABLED, OMNICACHE_MIN_SECRET_ENTROPY, OMNICACHE_PRESEEDED_KEYS, OPENAI_API_KEY and RAZORPAY_KEY_ID from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does omnicache-proxy run?
It speaks stdio and streamable-http, so it runs as a local process your client starts. It is published on npm as omnicache-ai at 3.1.0.
How current is this page?
The grade is for one exact copy of the source (997a04521553), read on 2026-10-11. The repository is watched and re-audited when it changes.