Skip to main content

Model Context Protocol

The Radar is an MCP.

Pipe the AI native engineering tools radar into Claude Desktop, Claude Code, Cursor, Windsurf, or any MCP-aware client. Eleven tools, one HTTP endpoint, no auth — query it like an LLM-native data layer.

Tools
11
Auth
None
Cost
Free
Transport
HTTP
POST https://radar.creative-technology.digital/api/mcp/mcp

Discovery file: /mcp.json · See the integration guides to wire it into your client.

Integration Guides

Claude Desktop

  1. Open claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\).
  2. Add the mcpServers block below (merge with existing if present).
  3. Save and restart the application to connect.

{
  "mcpServers": {
    "signal-radar": {
      "url": "https://radar.creative-technology.digital/api/mcp/mcp"
    }
  }
}

ChatGPT

  1. MCP support is in beta via custom GPT actions — there is no JSON config file like other native clients.
  2. In the GPT builder, go to Configure > Actions > Create new action.
  3. Set the server URL to the endpoint below and choose "No authentication" (the Signal MCP server requires none).

https://radar.creative-technology.digital/api/mcp/mcp

Windsurf

  1. Open your Codeium MCP config at ~/.codeium/[editor]/mcp_config.json (create it if it does not exist).
  2. Add the mcpServers block below (merge with existing if present).
  3. Reload the editor or run the MCP: Reload Servers command.

{
  "mcpServers": {
    "signal-radar": {
      "serverUrl": "https://radar.creative-technology.digital/api/mcp/mcp"
    }
  }
}

Claude Code

  1. For project-level config, create or edit .mcp.json in your project root.
  2. For user-level config, edit ~/.claude.json.
  3. Add the mcpServers block below (merge with existing if present).

{
  "mcpServers": {
    "signal-radar": {
      "type": "http",
      "url": "https://radar.creative-technology.digital/api/mcp/mcp"
    }
  }
}

Cursor

  1. Open Cursor → Settings → MCP (or edit ~/.cursor/mcp.json directly).
  2. Add the mcpServers block below (merge with existing if present).
  3. Restart Cursor — the radar tools will appear in the MCP tools list.

{
  "mcpServers": {
    "signal-radar": {
      "url": "https://radar.creative-technology.digital/api/mcp/mcp"
    }
  }
}

Any other MCP-compatible client: point it at https://radar.creative-technology.digital/api/mcp/mcp with JSON-RPC 2.0 over HTTP. No auth required.

Tools Library

* required

search_tools

Search Tools

Search the ACES-evaluated agentic tools radar. With a query, performs semantic search. With a view (risky/trending/recommended), returns that curated list — optionally filtered by a query. Results include ACES Signal Level (Validated/Assessed/Tracked/Detected) and Evidence Grade (A–D).

Reference

querystring

Natural language search query

viewenum

Curated view: all (default), risky, trending, recommended

severityenum

(view=risky) Filter by risk severity

lowGovernanceOnlyboolean

(view=risky) Only tools with compliance below 13/20 threshold

directionenum

(view=trending) Filter by trend direction

categorystring

(view=recommended) Filter by category name

signalLevelFilterenum

(view=recommended) Filter by ACES Signal Level

limitnumber

Maximum results to return (default: 5)

Examples

  • Search for code review tools
  • Find IDE tools with strong governance
  • Show critical-risk tools
  • Show trending tools moving up
  • Recommend adopted coding assistants

Try It Playground

Prompting Tips

Do

  • Use natural language queries that describe the capability you want.
  • Use view=risky/trending/recommended when you want a curated list without a search query.
  • Combine view and query to scope semantic search within a curated subset.

Don't

  • Do not use internal IDs when a plain-language query will do.
  • Do not treat an empty risky result as proof that all tools are equally safe.

get_tool_details

Get Tool Details

Get detailed ACES scores for a specific tool — the six dimensions, Rating (0–100), Signal Level (Validated/Assessed/Tracked/Detected), Evidence Grade (A–D), applied caps, and evaluation history.

Reference

toolIdstring

Exact tool ID

toolNamestring

Tool name (case-insensitive)

Examples

  • Get details for Cursor

Try It Playground

Prompting Tips

Do

  • Prefer toolId when you already have the exact identifier.
  • Use toolName for a quick follow-up after a search result.

Don't

  • Do not provide both toolId and toolName unless you need a fallback.

compare_tools

Compare Tools

Compare 2-3 tools side by side. With detail=false (default): dimensions and key differentiators. With detail=true: per-dimension rationale, win/loss tally, movement triggers, and follow-up questions.

Reference

toolIdsstring*

Comma-separated tool IDs (2-3)

detailboolean

When true, returns deep comparison with rationale, tally, and follow-ups

Examples

  • Compare Cursor and Copilot
  • Deep compare Cursor and Claude Code
  • Detailed comparison of GitHub Copilot, Cursor, and Cline

Try It Playground

Prompting Tips

Do

  • Compare two or three tools at a time for readable output.
  • Use detail=true when you need procurement-grade detail with rationale.
  • Pair with whats_changed when you want trajectories alongside the snapshot view.

Don't

  • Do not compare large groups of tools in a single request.
  • Do not use detail=true when a quick comparison would do.

get_tool_history

Get Tool History

Get version history and score changes for a specific tool. With detail=false (default): event list and totals. With detail=true: rating timeline, per-dimension deltas, cap timeline, and narrative summary.

Reference

toolIdstring

Exact tool ID

toolNamestring

Tool name (case-insensitive)

detailboolean

When true, returns deep analytics with rating timeline and narrative

fromstring

(detail=true) Inclusive start date filter (YYYY-MM-DD)

tostring

(detail=true) Inclusive end date filter (YYYY-MM-DD)

Examples

  • Show Cursor history
  • Tool history for Claude Code from 2026-01-01 to 2026-05-01

Try It Playground

Prompting Tips

Do

  • Use this after get_tool_details when you want change-over-time context.
  • Use detail=true when 'how did this tool change' matters more than 'what is its current state'.
  • Bound long windows with from/to to keep the narrative focused.

Don't

  • Do not pass both toolId and toolName — pick one.
  • Do not expect comparison output here; use compare_tools for side-by-side analysis.

get_timeline_events

Get Timeline Events

Browse the industry timeline of major events in agentic developer tools.

Reference

typeenum

Filter by event type

minSignificancenumber

Minimum significance level (1-5)

limitnumber

Max events to return

querystring

Keyword filter on title or description

Examples

  • Show major product launches with significance 4 or higher

Try It Playground

Prompting Tips

Do

  • Use minSignificance to keep the timeline focused on major changes.
  • Add a query when you are tracking a specific company or theme.

Don't

  • Do not ask for the entire timeline when a smaller slice will answer the question.

whats_changed

What's Changed

Return the 'what should I care about right now' view: promotions, demotions, top climbers/fallers, and tools that gained caps.

Reference

sincestring

Look back from this date (default: 7 days ago, YYYY-MM-DD)

Examples

  • What's changed this week
  • What's changed since 2026-04-01

Try It Playground

Prompting Tips

Do

  • Use this as a daily/weekly pulse check.
  • Pair with get_intel_briefings for narrative context behind the changes.

Don't

  • Do not assume zero changes means nothing happened — the radar updates monthly.

get_insights

Get Radar Insights

Get aggregate statistics about all ACES-evaluated tools on the radar — Rating averages, breakdown by Signal Level and category, and the top-scoring tools. Pass include=['caps'] to also receive the full scoring-cap taxonomy.

Reference

includestring

Optional extras: 'caps' adds the scoring-cap catalog to the response

capToolstring

When include=['caps'], scope cap results to a specific tool slug/ID/name

Examples

  • Summarize the current radar
  • List all scoring caps
  • Which caps are applied to Devin

Try It Playground

Prompting Tips

Do

  • Use this when you want a quick snapshot of the whole radar.
  • Pass include=['caps'] to understand why a tool is held back from a higher rating.
  • Cross-reference cap categories when assessing risk.

Don't

  • Do not expect detailed per-tool analysis in the aggregate view.
  • Do not assume removable=temporary; some caps are conditional.

search_intel

Search Intel

Unified search across the ai-intel source archive and weekly digests. Source filters (category, relevance, alignment, radarTool, timelineCandidatesOnly) apply to the source portion.

Reference

querystring

Keyword search query (searches sources and digests)

categorystring

Filter sources by category

relevanceenum

Filter sources by relevance level

alignmentstring

Filter sources by alignment field value

radarToolstring

Filter sources mentioning a specific radar tool name

timelineCandidatesOnlyboolean

Only return sources flagged as timeline candidates

limitnumber

Max results per category

Examples

  • Search intel for reasoning model launches
  • Find recent high-relevance coding assistant coverage

Try It Playground

Prompting Tips

Do

  • Use a focused phrase when you want sources and briefings on the same topic.
  • Start here when you are not sure whether to browse sources or digests.
  • Use relevance to surface the most strategic items first.

Don't

  • Do not treat this as a substitute for tool-specific radar queries.

get_intel_briefings

Get Intel Briefings

Get weekly intelligence digests. Without 'week': returns recent digest summaries. With week='latest' or 'YYYY-WXX': returns the full markdown of that briefing.

Reference

limitnumber

Number of recent digests to return when no 'week' is specified

weekstring

ISO week (e.g. '2026-W18') or 'latest' to get the full briefing markdown

Examples

  • Show the three latest weekly digests
  • Brief me on the latest week
  • Fetch the 2026-W18 weekly digest

Try It Playground

Prompting Tips

Do

  • Use week='latest' to get the full current briefing narrative.
  • Omit 'week' when you only need summaries of recent digests.
  • Use this to anchor longer responses with an authoritative weekly narrative.

Don't

  • Do not assume a draft briefing has been human-reviewed (look for the 'DRAFT' flag).
  • Do not expect raw source snippets here; use search_intel for that.

get_knowledge

Get Knowledge

Retrieve organizational knowledge by type: positions (AI stances), partnerships (vendor relationships), pursuits (business initiatives), or decisions (decision log).

Reference

typeenum*

Knowledge type to retrieve

statusstring

Filter by status (positions, partnerships, pursuits)

prioritystring

Filter by priority (pursuits only)

keywordstring

Keyword filter on title and body text

dateFromstring

Start date filter YYYY-MM-DD (decisions only)

dateTostring

End date filter YYYY-MM-DD (decisions only)

Examples

  • Show our positions on AI coding assistants
  • List active vendor partnerships
  • Show high-priority pursuits in progress
  • Show decisions made in Q2 2026

Try It Playground

Prompting Tips

Do

  • Start with type='positions' when you need our AI stances for a proposal.
  • Use type='decisions' with dateFrom/dateTo to audit a specific period.
  • Combine type with keyword for targeted lookups.

Don't

  • Do not confuse pursuits (business initiatives) with partnerships (vendor relationships).

submit_tool

Submit Tool

Submit a new agentic developer tool for evaluation. Creates a GitHub issue.

Reference

toolstring*

Tool name

productUrlstring*

URL to the product page

quickTakestring*

Brief description or elevator pitch

Examples

  • Submit a new evaluation candidate

Try It Playground

Prompting Tips

Do

  • Provide a clear product URL and a concise quick take.
  • Use this when the tool is not already present in the radar.

Don't

  • Do not include sensitive data in the submission fields.
  • Do not use placeholder URLs for real submissions.

How is this different from the website?

The website is the readable surface — radar charts, tool cards, briefing pages. The MCP is the same data, addressable by your LLM. Different jobs, same source of truth.

LLM-native
Claude, ChatGPT, and Cursor can call the radar like a function. No copy-paste, no scraping, no stale screenshots.
Scriptable
POST to one endpoint and you can build a CI gate, a Slack bot, a quarterly review draft — anything that can speak JSON-RPC.
Queryable
whats_changed, get_intel_briefings, get_knowledge — questions that took ten clicks on the site become a one-line prompt.

Discovery file: /mcp.json · Endpoint: https://radar.creative-technology.digital/api/mcp/mcp · No auth, no rate limits today.