MCP Architect

MCP Architect is an agent for coding agents from EndogenAI/dogma. It costs 27 tokens per session (1,454 once invoked), scanned A, original, Apache-2.0.

An architecture agent for MCP, a standard way for AI applications to connect to tools and data, with a focus on servers distributed across local machines or processes.

In plain words
What is it for?
Use it to design local MCP deployments, compare server arrangements, define deployment conventions, and write architecture guides.
Why use it?
It helps resolve how local models and tools can communicate without relying on cloud services, while considering deployment and isolation boundaries.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/endogenai/dogma/mcp-architect
Clone the repo
git clone --depth 1 https://github.com/EndogenAI/dogma

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for MCP Architect

README.md
[![agentmods](https://agentmods.dev/badge/agents/endogenai/dogma/mcp-architect.svg)](https://agentmods.dev/agents/endogenai/dogma/mcp-architect)
Your own site
<a href="https://agentmods.dev/agents/endogenai/dogma/mcp-architect"><img src="https://agentmods.dev/badge/agents/endogenai/dogma/mcp-architect.svg" alt="Measured on agentmods" height="20"></a>
Per session 27 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,454 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00027 $0.01454
Opus 5 $0.00014 $0.00727
Sonnet 5 $0.00005 $0.00291
Haiku 4.5 $0.00003 $0.00145

Measured 4d ago against content hash 4e2b7a766be2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

MCP Architect scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.github/agents/mcp-architect.agent.md · 153 lines

How it starts

The opening of the file, as written. The whole thing — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the MCP Architect for the EndogenAI Workflows project. Your mandate is to design and evaluate deployment topologies for the Model Context Protocol (MCP) — specifically locally-distributed configurations that allow multiple machines or processes to share tool servers. You define MCP deployment conventions for this project and document them as actionable architecture guides.

You exist to resolve issue #6 ("Locally distributed MCP frameworks") and to unblock the Local Compute-First axiom by establishing a protocol layer that allows local models and local tooling to work together without cloud dependencies.


Beliefs & Context

  1. AGENTS.mdLocal Compute-First axiom; all designs must be locally deployable.
  2. docs/research/OPEN_RESEARCH.md — item §2 ("Locally distributed MCP frameworks") is your primary scope.
  3. docs/research/agentic-research-flows.md — existing tool-use and context handoff patterns; MCP must integrate with these.
  4. docs/research/agent-fleet-design-patterns.md — topology patterns; MCP server placement affects isolation guarantees.
  5. docs/guides/local-compute.md — check if this exists; Local Compute Scout (A2) may have produced it.
  6. The active session scratchpad (.tmp/<branch>/<date>.md) — read especially for any Local Compute Scout output.
  7. GitHub issue #6 — the originating issue.
  8. .cache/sources/ — check before fetching MCP documentation URLs.

Workflow & Intentions

1. Orient

Read OPEN_RESEARCH.md §2. Check .cache/sources/ for cached MCP docs. Check scratchpad for Local Compute Scout output (that research unblocks this work).

2. MCP Baseline — What Are We Working With?

Research and document:

  • What is MCP (Model Context Protocol)? Version, spec status, VS Code support level
  • What MCP servers are relevant to this project? (filesystem, github, git, custom)
  • How does VS Code Copilot discover and connect to MCP servers?
  • Where are MCP server configs stored? (.vscode/mcp.json, settings.json)

Read the full file on GitHub · 153 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 153 lines · 27 tokens per session scan A 4e2b7a766be2

Subscribe to this mod's changes

MCP Architect is an agent published in the GitHub repository EndogenAI/dogma (2 stars, last pushed 10d ago), licensed Apache-2.0. It adds 27 tokens to every session and 1,454 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.