Brunel

Agents and API

An infrastructure architect for your AI agents

Brunel speaks the Model Context Protocol and has a REST API. Your coding agent can analyze a repository, design priced infrastructure, refine it and fetch the Terraform, with the same guarantees as the app.

Connect an agent REST API
  • Claude Code
  • Claude Desktop
  • Cursor
  • VS Code
  • Windsurf
  • Any MCP client
terminal · Claude Code
$ claude mcp add --transport http brunel \
    https://brunel.cloud/api/mcp \
    --header "Authorization: Bearer $BRUNEL_KEY"

> Design the production infrastructure of
  this repository for 2,000 users, €300/month

● brunel · analyze_repository
● brunel · design_infrastructure
● brunel · get_files  tier=growth

How it works

Three steps to connect

No SDK, no plugin to install: a URL and a key.

1. Create a key

In the console, create a read-only or read-and-write key, with an expiry. It is shown once and stored as a hash.

2. Add the server

Paste the configuration for your client. The console gives it for Claude Code, Claude Desktop, Cursor, VS Code and Windsurf.

3. Ask in plain words

The agent calls Brunel's tools. Every project also appears in Brunel with its diagram, costs and files.

MCP tools

7 tools, one flow

Analyze, design, refine, fetch. Read-only keys can only call the read tools.

  • whoamiread only

    The Brunel account behind this key: login, plan and the key's scope.

  • list_projectsread only

    The account's analyses (one per repository conversation), most recent first.

  • analyze_repositoryread and write

    Read a GitHub repository and return what it runs (languages, frameworks, datastores, queues…) with file evidence, its services, a security check and a suggested context for design_infrastructure. Takes 5 to 60 seconds.

  • design_infrastructureread and write

    Design four priced production tiers (starter, growth, scale, enterprise) for an analyzed repository. Every field of the context is optional: missing ones come from the suggested context.

  • get_architectureread only

    The current state of a project: facts, security check and, once designed, the four priced tiers with their components.

  • refine_architectureread and write

    Ask a question about the architecture or request a change ("make it cheaper", "add a staging environment", "switch to Scaleway"). A change creates a new version; the answer and the new tiers are returned.

  • get_filesread only

    The deployable files of one tier: Terraform, Kubernetes or Compose, CI workflow, runbook and decision records. Pass a path to get one file's content.

REST API

The same actions, over HTTPS

For scripts, CI jobs and your own products: JSON in, JSON out. Analysis and design calls wait for the result, typically a few seconds.

  • POST/api/v1/projects· analyze a repository
  • POST/api/v1/projects/:id/design· four priced tiers
  • POST/api/v1/projects/:id/messages· ask or change
  • GET/api/v1/projects/:id/files· Terraform, CI, docs
Full reference in the console
shell
curl -X POST https://brunel.cloud/api/v1/projects \
  -H "Authorization: Bearer $BRUNEL_KEY" \
  -H "Content-Type: application/json" \
  -d '{"repoUrl":"https://github.com/acme/shop"}'

{ "data": {
    "conversationId": "k3v9x2…",
    "status": "awaiting_context",
    "facts": [ { "key": "postgres",
                 "evidence": ["docker-compose.yml:12"] } ],
    "suggestedContext": { "monthlyBudgetEur": 100 }
} }

Guarantees

Agents get the same rules as people

Your plan, quotas, team IP allowlist and access checks apply to every call. Keys are hashed, can expire, and can never manage keys, your account, billing or your team's security.

Connect your first agent

Create a key and paste one line in Claude Code. The console fills in the configuration for you.

Open the console Take the product tour
Agents and API · Brunel