MCP Setup

Connect AI agents to DataHippo using the Model Context Protocol. Configure your MCP client to connect to the DataHippo API with an agent token.

What this unlocks

After completing setup, your agent can list projects, query traces and logs, manage prompt configurations, run playground comparisons, create dashboards and alert rules, and access all 100+ platform operations through 6 high-level tools.

Prerequisites

• A DataHippo project with an agent token (create one under Settings → Agent Tokens in your project)

SDK keys are not accepted

MCP requires an agent token. Regular SDK API keys will be rejected with a 403 error. Agent tokens provide scoped access specifically designed for AI agent integrations — they can be individually revoked and are tracked separately in audit logs.

Cursor

Add the following to your Cursor MCP configuration (.cursor/mcp.json in your project root or ~/.cursor/mcp.json for global settings):

{
  "mcpServers": {
    "datahippo": {
      "url": "https://datahippohq.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${env:DATAHIPPO_AGENT_TOKEN}"
      }
    }
  }
}

Then set the environment variable with your agent token:

export DATAHIPPO_AGENT_TOKEN="dh_agent_..."

Add the export to your shell profile (~/.zshrc, ~/.bashrc, etc.) so it persists across sessions. Restart Cursor after making changes.

TIP

Cursor uses ${env:VAR_NAME} syntax to interpolate environment variables in MCP config. This keeps tokens out of version control.

Claude Code

Add DataHippo for your current project:

claude mcp add --transport http --header "Authorization: Bearer $DATAHIPPO_AGENT_TOKEN" datahippo \
  https://datahippohq.com/api/mcp

Or make it available across all your projects:

claude mcp add --transport http --scope user --header "Authorization: Bearer $DATAHIPPO_AGENT_TOKEN" datahippo \
  https://datahippohq.com/api/mcp

Set the environment variable in your shell profile (~/.zshrc, ~/.bashrc, etc.) before running the command:

export DATAHIPPO_AGENT_TOKEN="dh_agent_..."

To share the config with your team via .mcp.json, use the mcp-remote bridge (requires Node.js):

{
  "mcpServers": {
    "datahippo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://datahippohq.com/api/mcp",
        "--header",
        "Authorization:Bearer ${DATAHIPPO_AGENT_TOKEN}"
      ]
    }
  }
}

Since Claude Code inherits your shell environment, ${DATAHIPPO_AGENT_TOKEN} is resolved at runtime from the export above.

Do not paste tokens directly into config files

MCP config files are readable by every agent and tool that has access to your project. If a token appears as a literal value in the file, any MCP server — not just DataHippo — can read it. Always use environment variable references so the secret stays in your shell profile.

Claude Desktop

Claude Desktop does not natively support remote HTTP servers with custom headers. Use a wrapper script with the mcp-remote bridge instead.

1. Set the environment variable

Add the export to your shell profile (~/.zshrc, ~/.bashrc, etc.):

export DATAHIPPO_AGENT_TOKEN="dh_agent_..."

2. Create a wrapper script

Claude Desktop is a GUI app and does not inherit shell environment variables. A small wrapper script sources your profile so the token is available:

cat > ~/.local/bin/datahippo-mcp.sh << 'EOF'
#!/bin/sh
. "$HOME/.zshrc" 2>/dev/null || . "$HOME/.bashrc" 2>/dev/null
exec npx -y mcp-remote \
  https://datahippohq.com/api/mcp \
  --header "Authorization:Bearer ${DATAHIPPO_AGENT_TOKEN}"
EOF
chmod +x ~/.local/bin/datahippo-mcp.sh

3. Add to Claude Desktop config

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "datahippo": {
      "command": "sh",
      "args": ["-c", "~/.local/bin/datahippo-mcp.sh"]
    }
  }
}

Restart Claude Desktop. You should see DataHippo's tools available in the tool picker.

Other MCP Clients

Any MCP client that supports Streamable HTTP transport can connect. Point it at:

Setting	Value
URL	https://datahippohq.com/api/mcp
Auth	Authorization: Bearer <your-agent-token>

Scoped Permissions

Agent tokens have configurable scopes that control which tools the agent can access. Tools requiring a scope the token doesn't have are hidden entirely.

By default, new agent tokens are created with read-only scopes. A write scope implicitly grants the corresponding read scope — for example, llm:write also grants llm:read.

To grant an agent write access, include the write scopes when creating the agent token in project settings.

General

project:read

View projects, API keys, and project statistics.

"List my projects" · "Show the API keys for this project" · "What are my exception stats?"

project:write

Create and update projects, generate new API keys.

"Create a new project called mobile-app" · "Generate an API key for the staging project"

billing:read

View billing usage, per-project breakdowns, forecasts, and budget status.

"What's my current usage?" · "Show the usage forecast for this billing period" · "Am I near my budget limit?"

Flow (Prompt Hub & LLM Gateway)

llm:read

View provider integrations, gateway settings, prompt configs, sessions, metrics, playground results, and evaluation scores.

"What providers are configured?" · "Show me LLM cost for the last 7 days" · "List prompt configs" · "What's the latency breakdown by model?"

llm:write

Configure provider integrations (add/rotate API keys), update gateway settings, create and deploy prompt versions, manage rollouts, submit evaluation scores.

"Add an Anthropic integration with this API key" · "Deploy prompt config customer-support to version 3" · "Enable PII masking in the guardrails" · "Rollback the current rollout"

Watch (APM)

observability:read

View exceptions, traces, logs, incidents, dashboards, alert rules, fired alerts, health checks, profiling data, API endpoints, services, maintenance windows, and notification channels.

"Show me the top exceptions this week" · "Get trace abc-123" · "Search logs for 'timeout'" · "Which alert rules have fired?" · "List health checks"

observability:write

Create and modify dashboards, widgets, alert rules, notification channels (Slack, Teams, Discord, PagerDuty, ServiceNow, webhooks), health checks, maintenance windows, and cloud integrations (AWS, Azure, GCP, OCI). Link GitHub repos.

"Create a dashboard from the API monitoring template" · "Add a Slack notification channel" · "Create an alert rule for error rate > 5%" · "Set up a maintenance window for Saturday 2-4am"

Verifying the Connection

Once connected, try asking your agent:

• "List my DataHippo projects"
• "Show me the LLM gateway overview for the last 7 days"
• "What alert rules are configured?"

Next Steps

• Available Tools — full reference of all 6 tools and their operations
• In-App Agent — use the built-in AI assistant in the DataHippo UI