Skip to main content
MCP server CX Codex

pAIchart MCP Hub — Claude Code MCP server for infrastructure design review

pAIchart MCP Hub turns requirements and network topology into a reviewed low-level design: per-device config, verification commands, rollback. Does not apply anything.

no license updated 2026-10-24

Install into Claude Code · Platform CLI: remote server All methods and platforms ↓
claude mcp add --transport http mcp-hub https://paichart.app/mcp
Open on GitHub

What pAIchart MCP Hub does

The server runs a directed acyclic graph of agents, each responsible for one slice: network devices, Terraform, Kubernetes, or the observability stack. One agent captures live state, another designs, a third writes config, a fourth reviews. Edges in the graph carry values that only exist after a leg runs, so the cloud leg authorises the exact address range the network leg derived.

What you get back is a change package: config for each device, commands to prove it worked, and a rollback path. Your team applies it out of band, idempotently. The tool designs and reviews changes; it never touches production itself.

Correctness checks come in three layers. The first is arithmetic in code rather than an LLM's opinion: a CIDR range must cover exactly its declared members, an AS number must match what the live state allows. The second layer is independent reviewers, one per domain plus one across all of them. The third is a release gate where everything meets in a deterministic AND. There is no confidence score anywhere in it.

It runs from https://paichart.app/mcp and reads device state through per-device MCP servers. You need a requirements.md and topology.json at a fetchable location, otherwise there is nothing to work from.

What it helps with

  • stop hand-reconciling values across infrastructure domains
  • get a reviewed low-level design instead of a junior draft
  • catch a CIDR range that swallows already-allocated addresses
  • see which checks could not run at all
  • confirm a change does not touch anything beyond its scope

Example prompts for the agent

  • Build a low-level design from topology.json and the requirements in requirements.md.
  • Check that the derived CIDR range does not sweep in already-used addresses.

How to install pAIchart MCP Hub

Platform CLI: remote server

claude mcp add --transport http mcp-hub https://paichart.app/mcp

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about pAIchart MCP Hub

How do I install pAIchart MCP Hub in Claude Code?

Run claude mcp add --transport http mcp-hub https://paichart.app/mcp. Alternatively, with the name paichart-mcp-hub: claude mcp add --transport http paichart-mcp-hub https://paichart.app/mcp.

What does pAIchart MCP Hub do?

From requirements.md and topology.json it produces a reviewed low-level design: per-device config, verification commands, rollback. It works as a graph of agents across network, Terraform, Kubernetes and observability domains.

How does it check changes for correctness?

Three layers. First, computations in code, not model judgement. Second, independent reviewers, one per domain and one integration reviewer across all. Third, a release gate that combines every condition with AND.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open