MCP Server Setup

Connect your AI coding agent to FlameOn for automatic performance analysis.

Overview

FlameOn exposes an MCP (Model Context Protocol) server that your coding agent can query to find performance issues, compare releases, detect N+1 queries, and more — all without leaving your editor or terminal.

Your MCP endpoint is:

https://flameon.ai/mcp

You'll need an API key for authentication. Generate one from your application's settings page under Teams → App → Settings → Application API Keys.

Every request must also include an X-Application header that identifies which team and app to query, in the format:

X-Application: your-team-slug/your-app-slug

The team and app slugs are the same values that appear in your dashboard URL: /teams/your-team-slug/apps/your-app-slug

Claude Code

Claude Code supports MCP servers natively via the CLI or a project config file.

Option 1: CLI command

claude mcp add --transport http flameon \
  --header "Authorization: Bearer YOUR_TOKEN" \
  --header "X-Application: your-team-slug/your-app-slug" \
  https://flameon.ai/mcp

Option 2: Project config (.mcp.json)

Create .mcp.json in your project root:

{
  "mcpServers": {
    "flameon": {
      "type": "http",
      "url": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

Use ${FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

OpenAI Codex

Codex uses a TOML config file with native streamable HTTP support.

Edit ~/.codex/config.toml

[mcp_servers.flameon]
url = "https://flameon.ai/mcp"
bearer_token_env_var = "FLAMEON_API_KEY"

[mcp_servers.flameon.headers]
X-Application = "your-team-slug/your-app-slug"

Then set the environment variable:

export FLAMEON_API_KEY="your-api-key-here"

Or via CLI

codex mcp add flameon --url https://flameon.ai/mcp

Then edit ~/.codex/config.toml to add bearer_token_env_var and the X-Application header.

Gemini CLI

Gemini CLI supports streamable HTTP via the httpUrl field.

Option 1: CLI command

gemini mcp add --transport http flameon \
  https://flameon.ai/mcp \
  --header "Authorization: Bearer YOUR_TOKEN" \
  --header "X-Application: your-team-slug/your-app-slug"

Option 2: Edit ~/.gemini/settings.json

{
  "mcpServers": {
    "flameon": {
      "httpUrl": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

Use ${FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

Amp

Amp (by Sourcegraph) uses streamable HTTP by default for remote MCP servers.

Edit ~/.config/amp/settings.json

{
  "amp.mcpServers": {
    "flameon": {
      "url": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

For project-scoped config, create .amp/settings.json in your project root (requires approval on first run).

Use ${FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

OpenCode

OpenCode supports remote MCP servers with custom headers.

Edit opencode.json

Place in your project root or ~/.config/opencode/opencode.json for global config:

{
  "mcp": {
    "flameon": {
      "type": "remote",
      "url": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer {env:FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      },
      "oauth": false,
      "enabled": true
    }
  }
}

Use {env:FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

Cursor

Cursor auto-detects streamable HTTP when you provide a url field.

Edit .cursor/mcp.json

Create in your project root or ~/.cursor/mcp.json for global:

{
  "mcpServers": {
    "flameon": {
      "url": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${env:FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

Use ${env:FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

Windsurf

Windsurf uses serverUrl for HTTP-based MCP servers.

Edit ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "flameon": {
      "serverUrl": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${env:FLAMEON_API_KEY}",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

Use ${env:FLAMEON_API_KEY} in place of the token value to read from the FLAMEON_API_KEY environment variable.

Zed

Zed configures MCP servers under context_servers in its settings file.

Edit ~/.config/zed/settings.json

{
  "context_servers": {
    "flameon": {
      "url": "https://flameon.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN",
        "X-Application": "your-team-slug/your-app-slug"
      }
    }
  }
}

If your build of Zed doesn't support native HTTP yet, use mcp-remote as a proxy:

{
  "context_servers": {
    "flameon": {
      "command": {
        "path": "npx",
        "args": [
          "-y", "mcp-remote",
          "https://flameon.ai/mcp",
          "--header",
          "Authorization:Bearer YOUR_TOKEN",
          "--header",
          "X-Application:your-team-slug/your-app-slug"
        ]
      }
    }
  }
}
Generic / Other Agents

FlameOn's MCP server implements the MCP Streamable HTTP transport (spec version 2025-03-26). Any MCP-compatible client can connect.

Connection details

Endpoint https://flameon.ai/mcp
Transport Streamable HTTP (POST)
Auth header Authorization: Bearer YOUR_TOKEN
App header X-Application: team-slug/app-slug
Content type application/json

Using mcp-remote as a stdio bridge

If your agent only supports stdio-based MCP servers, use mcp-remote to bridge:

npx -y mcp-remote https://flameon.ai/mcp \
  --header "Authorization: Bearer YOUR_TOKEN" \
  --header "X-Application: your-team-slug/your-app-slug"

Point your agent's MCP config at this command using its stdio/local transport type.

Available Tools

Once connected, your agent will have access to these tools:

Tool Description
investigate_event_performance Guided investigation with hotspots, traces, and next steps
list_events List profiled events with p50/p95/p99 latency
list_traces_for_event List concrete trace IDs for an event
get_profile Hot functions for an event, ranked by total time
get_call_tree Full call hierarchy with timing (flame graph data)
get_trace_profile Single-trace profile and samples for outlier inspection
get_call_tree_for_trace Call tree for one concrete trace
get_function_detail Cross-event detail for a specific function
detect_anomalies Auto-detect N+1 queries and slow events
get_slow_traces Outlier traces for an event
compare_periods Compare two time periods for an event
diff_functions_between_periods Function-level time-window diff for a specific event
FlameOn

Continuous profiling and flamegraph analysis for production applications.