AI Agent Integration (MCP)

Axiom natively supports the Model Context Protocol (MCP), an open standard that allows logic-driven AI agents (like Claude Desktop, Cursor, or custom scripts) to securely interface with your decision matrix.

By connecting an agent via MCP, you delegate heavy contextual reasoning to an LLM. It can autonomously evaluate options, suggest criteria modifications, and post scores back to Axiom, effectively serving as a synthetic Contributor on your project.

Server Capabilities

Our MCP server exposes a robust suite of tools that your agent can invoke:

• State Ingestion:
  • mcp_axiom_list_projects: Discover available matrices.
  • mcp_axiom_get_project_context: Pull down the entire schema of a project, including all active criteria and current options.
• Structural Mutation:
  • mcp_axiom_add_criteria: Define new scoring or elimination constraints.
  • mcp_axiom_propose_option: Discover a new candidate and add it to the matrix.
• Evaluation & Scoring:
  • mcp_axiom_score_option: Synthesize external data and post a numerical score for a specific option against a specific criterion.
  • mcp_axiom_toggle_elimination: Autonomously eliminate an option if it fails a dealbreaker (e.g., parsing a vendor's security docs and eliminating them for lacking SOC2).

Connecting Claude Desktop

1. Provisioning an API Key

To grant an agent network authority, you need an API key bounded to your Organization.

1. Navigate to Organization Settings.
2. Drill into the API Keys sub-section.
3. Mint a new key. Store this securely—it carries the same RBAC privileges as an Org Admin.

2. Client Payload Configuration

To inject the MCP server into your local client, append the following JSON payload to your client's core integration file (e.g., claude_desktop_config.json).

Replace YOUR_API_KEY_HERE with your minted key.

{
  "mcpServers": {
    "axiom": {
      "command": "npx",
      "args": ["-y", "axiom-agent-mcp@latest"],
      "env": {
        "AXIOM_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Advanced Prompting Strategies

Once connected, you can build powerful autonomous workflows. Here are a few advanced prompts to try with your agent:

The "Auto-Scorer"

"Pull my 'Marketing CRM Selection' project in Axiom. Search the web for the pricing pages of all the current options. Evaluate them against my 'Pricing & Budget' criteria. Use the mcp_axiom_score_option tool to post the scores back to Axiom with a detailed rationale."

The "Ruthless Eliminator"

"Read the attached PDF of Vendor Beta's security compliance report. Check it against the Elimination criteria in my Axiom project. If they do not explicitly state they have HIPAA compliance, use mcp_axiom_toggle_elimination to instantly disqualify them and explain why."