MCP integration

Aqmen ships a built-in MCP server (Model Context Protocol), so any MCP-capable agent can work inside your projects: structure analyses, research and populate cited values, apply hypotheses, and run scenarios. This is the agent-first side of the platform: the UI is always there, but an agent moves faster and leaves the same reviewable trail.

Authorization is OAuth with your Aqmen account. The agent signs in as you, works inside your organization's workspace with the same access you have, and there are no API keys to create, paste, or rotate.

Your MCP endpoint:

https://app.aqmen.ai/mcp

Claude

1. In Claude (web or desktop), open Settings → Connectors → Add custom connector.
2. Enter the Aqmen MCP URL: https://app.aqmen.ai/mcp.
3. Claude opens Aqmen's sign-in. Authorize with your Aqmen account.

Once connected, enable the connector in a conversation and ask Claude to list your projects to confirm everything works.

ChatGPT

In ChatGPT, add a connector (under Settings → Connectors, or developer mode depending on your plan) and point it at the same URL: https://app.aqmen.ai/mcp. OAuth works the same way: ChatGPT sends you to Aqmen's sign-in to authorize.

Claude Code and other clients

Any MCP client works: point it at https://app.aqmen.ai/mcp. Aqmen supports OAuth with dynamic client registration, so clients register themselves; no manual credential setup. In Claude Code, for example:

claude mcp add --transport http aqmen https://app.aqmen.ai/mcp

The first request opens the OAuth sign-in in your browser.

What an agent can do once connected

• Orient: read your projects, analyses, existing hypotheses, and sources before touching anything. Informed, not blind.
• Structure: draft a market driver tree, propose competitive-landscape fields, scaffold a company P&L.
• Populate: research values and land them with sources attached, never bare numbers.
• Forecast: create quantitative hypotheses, apply them to drivers, and dial them per scenario.
• Iterate: re-cut a tree, split a segment, adjust an assumption, and report exactly what changed.

Everything lands as real, reviewable artifacts in the project (structures, values, assertions), not text in a chat transcript. You review the agent's work the same way you'd review a colleague's.

Good first prompts

• "List my Aqmen projects and summarize what's in each."
• "In the [target] project, size the EU industrial-software market bottom-up (geography × end-market) and cite every value."
• "Benchmark the target's top 8 competitors on pricing model and NPS."
• "Apply a downside scenario where pricing stays flat and adoption slips two years. What changes in 2030?"

Guardrails

The agent acts as your signed-in user, so its work is always attributable. Values carry their sources, forecasts read as named hypotheses rather than buried formulas, and nothing counts as reviewed until a human confirms it. Disconnect any time from your client's connector settings: access rides on your Aqmen session, not on stored keys.