TN-MCPs
Provides policy-checked, audited, approval-gated access to Cloudflare infrastructure through a dedicated Cloudflare MCP server, wrapping Cloudflare's official APIs and MCP server for managing Cloudflare resources such as ingress and domain policies.
Provides policy-checked, audited, approval-gated access to GitHub through a GitHub App, enabling interaction with GitHub repositories and related operations via the GitHub API.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@TN-MCPsshow me recent Cloudflare audit events"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
TN-MCPs
Telos Nexus MCP platform. The monorepo for the Telos Nexus Model Context Protocol control plane: separate MCP servers per provider behind a thin authenticated gateway. Through it, AI clients (Claude Code, Codex/OpenAI clients, JARVIS, future agents) get policy-checked, audited, approval-gated access to Telos Nexus infrastructure. Cloudflare comes first, then GitHub, Oracle, and Telos workflow tools.
AI clients ──HTTPS──▶ mcp.telosnexus.cloud/<provider>/mcp
│ Cloudflare Access (one path-scoped app per provider)
▼
Cloudflare Tunnel (dedicated)
│
A1 VM, loopback only ▼
gateway :8790 Host/Origin · routing · request IDs · rate limits · JWT check · approvals
├─▶ mcps/cloudflare :8701 Cloudflare credentials only
├─▶ mcps/github :8702 GitHub App only (later)
├─▶ mcps/oracle :8703 OCI, allowlisted ops only (later)
└─▶ mcps/telos-control :8704 semantic Telos workflows (later)
shared: auth · policy (R0–R3, scopes, domain registry) · audit · approvals · config · observabilityTN-MCPs does not re-implement provider APIs. It wraps official APIs and official MCP servers in a Telos Nexus policy, audit, and approval layer. Models are replaceable clients. The durable asset is TN-MCPs plus Telos policy.
Status
Phase 1 (of phases 0–15): repository foundation. Nothing is deployed and
mcp.telosnexus.cloud does not exist yet. The owner's architecture and security
contract is docs/TN_MCPS_MASTER_IMPLEMENTATION_PLAN.md.
The phased execution plan that implements it, including the deviations awaiting owner
sign-off, is docs/MASTER_PLAN.md.
Related MCP server: Pincer-MCP
Stack
Runtime | Node.js 24 LTS |
Language | TypeScript 7 (strict, ESM, NodeNext) |
MCP | Official TypeScript SDK v2 ( |
Workspace | pnpm 12 workspaces ( |
Lint / format | Biome 2 |
Tests | Vitest 5 |
Host | Oracle Cloud A1 (arm64), native systemd services, Cloudflare Tunnel ingress, no open inbound ports |
The reasoning behind each choice is in docs/ARCHITECTURE.md.
Develop
pnpm install # uses the committed lockfile; dependency build scripts stay blocked
pnpm run check # THE gate: biome ci + tsc -b + vitest + secret guard
pnpm test # tests only
pnpm run format # apply formattingRequires Node 24 (.node-version) and pnpm 12 (packageManager in package.json).
Layout
packages/shared/ redaction + typed config loading (Phase 1)
scripts/ guard-secrets.mjs, the repo-local secret gate
docs/ owner contract, master plan, architecture, security, deployment, Cloudflare, Claude Code
.github/ CI on GitHub-hosted runners (read-only token), DependabotEach phase creates the directories it needs: gateway/, mcps/cloudflare/,
packages/mcp-common/, packages/policy/, deploy/, and so on. See ARCHITECTURE §10.
Documentation
docs/TN_MCPS_MASTER_IMPLEMENTATION_PLAN.md: the owner's architecture and security contract
docs/MASTER_PLAN.md: phased execution plan, contract mapping, deviations, owner actions
docs/ARCHITECTURE.md: system design and decisions
docs/SECURITY.md: rules, threat model, credential isolation, R0–R3
docs/DEPLOYMENT_A1.md: the Oracle A1 host and how TN-MCPs will run on it
docs/CLOUDFLARE.md: Cloudflare ingress, the Cloudflare MCP, the domain policy
docs/CLAUDE_CODE.md: connecting Claude Code
CLAUDE.md: instructions for AI engineers working in this repo
Security
This repository is public. It contains no credentials and never will. Credentials live on the host in per-process permissioned files and are documented by location only (docs/SECURITY.md §3). If you believe you have found a secret or a vulnerability here, do not open a public issue. Contact the repository owner directly through GitHub.
This server cannot be deployed
Maintenance
Related MCP Connectors
Give AI agents identity, scoped access, trusted context, and verifiable actions through MCP.
Connect MCP clients to 2,000+ AI models without managing provider API keys.
Zero-secret MCP gateway for AI agents: risk-scored, audited calls with human-in-the-loop approval.
Zero-setup MCP gateway securely connecting AI to your tools with authentication and workflows
Related MCP Servers
- FlicenseNot gradedqualityAmaintenanceA production-ready MCP gateway and control plane that provides credential vault, policy engine, audit logging, and managed runtime for routing tool calls between AI agents and downstream MCP servers.58-
- AlicenseNot gradedqualityCmaintenanceA security-hardened MCP gateway that enables AI agents to call LLM APIs (Gemini, OpenAI, Claude, etc.) using ephemeral proxy tokens, eliminating exposure of real API keys.37 npm6Apache 2.0
- AlicenseNot gradedqualityBmaintenanceProvides a secure MCP gateway for AI agents to access APIs without exposing raw credentials, with scoped access, audit logging, and OAuth support.MIT
- AlicenseNot gradedqualityBmaintenanceProvides an isolated MCP gateway for SynapXnet AIOps, DataOps, and MLOps evidence-to-remediation workflows, with OAuth validation, scoped tool discovery, persistent approvals, and audit tracking.AGPL 3.0