Intel Hub

Live, cited intelligence for your AI agent. One command to install.

MCP server exposing 5 free tools (search, feed, knowledge graph, read, daily briefings) across domain-specific feeds (Claude Code, Solana, AI video, and more). Claim a free key to use intel_ask for synthesized answers with citations.

1 Install

Installing for agentic-x. The MCP server label will be intel-agentic-x — install multiple packages alongside the generic intel install.

Install in your editor. Adds intel-agentic-x as a labeled MCP server so you know which install corresponds to agentic-x.

Add to Cursor Add to VS CodeAdd to VS Code Insiders
or run in your terminal
claude mcp add --transport http --scope user intel https://intel.lemuriaos.ai/mcp

Restart Claude Code. Done. Already in a Claude Code session? Paste this prompt and the agent installs it for you:

Install the Intel Hub MCP server as "intel-agentic-x" by running: claude mcp add --transport http --scope user intel https://intel.lemuriaos.ai/mcp

Agent note: Run the command verbatim. Do not hand-author JSON in ~/.claude/.mcp.json — that triggers dotfile-overwrite security scanners in some agents. The claude mcp add subcommand is the supported path.

Hermes, Continue, or other MCP clients — manual config
{ "mcpServers": { "intel": { "type": "url", "url": "https://intel.lemuriaos.ai/mcp" } } }

Paste into ~/.cursor/mcp.json (Cursor manual path) or ~/.claude/.mcp.json (Claude Code manual path). Restart the editor after.

Hermes Agent: Use the CLI subcommand — hermes mcp add intel-agentic-x --url https://intel.lemuriaos.ai/mcp — when prompted, pick header auth and set Authorization: Bearer <your-key>. Or add the YAML block above under mcp_servers: in ~/.hermes/config.yaml.

2 Try it

Once your editor sees the MCP server, ask your agent one of these:

  • Use intel_briefing to get today's briefing→ Today's executive summary across tracked packages with inline [Source] citations. Free.
  • Use intel_search to find the most-cited update in the last 48h→ Ranked article/tweet/release list with impact scores and strategic dimensions. Free.
  • Use intel_knowledge to get verified facts about a company or model→ Structured facts (funding, benchmarks, catalysts) with confidence scores and freshness decay. Free.
  • Use intel_ask to summarize what changed this weekkey→ Synthesized natural-language answer pulling from the 7-day rolling feed. Requires a claimed key — free, 30 seconds.

Available Tools

5 tools free, no key. intel_ask (synthesized answers) requires a claimed key — claim one below, free, 30 seconds.

intel_askkey

Ask a question about intelligence from monitored sources. Returns a synthesized answer with inline [Source Name] citations, backed by verified knowledge.

intel_search

Search for articles, tweets, papers, and releases. Returns enriched results with summaries, knowledge items, and strategic impact scores.

intel_read

Get full details for a specific item: complete content, extracted entities, structured knowledge (metrics, catalysts, techniques), and strategic impact assessment. Use after intel_search to dive deeper.

intel_feed

Get a topic-filtered feed or a synthesized digest briefing. Use package param for topic-filtered feed, or set digest:true for a daily briefing.

intel_knowledge

Query the structured knowledge graph for verified facts: funding amounts, benchmarks, metrics, catalysts, techniques. Each fact has a confidence score (0-1) and freshness decay.

intel_briefing

Get the latest daily or weekly intelligence briefing. `package` is optional: omit to get a catalog listing (name + description + latest executive summary for every package) so you can pick; pass a specific package name for that package's full briefing with the So-What callout.

Need more than 10 requests per minute? Claim a key

A claimed key bumps your rate limit to 30 requests per minute and scopes to a specific package. Free.

Select your package:

Loading packages...

Troubleshooting

Do I need a key?

No. Anonymous use gets 10 requests per minute per IP, no signup. Shared IPs (coworking, café Wi-Fi) pool the quota across everyone on that network. Claim a key above to upgrade to 30/min per key, scoped to a specific package.

“MCP server not found”

If you used the claude mcp add command, run claude mcp list to confirm it registered. For manual config, the JSON goes in ~/.claude/.mcp.json (Claude Code) or ~/.cursor/mcp.json (Cursor). Restart the editor after.

“Unauthorized” or “Invalid API key”

Double-check your key starts with ik_ and was copied completely. Keys expire after 90 days — re-claim above to get a fresh one.

“Rate limited”

Anonymous tier: 10/min per IP. Claimed-key tier: 30/min per key. Limits use a sliding 60-second window — wait one minute and try again, or claim a key above if you hit the anon limit consistently.

Need help?

Email support@lemuriaos.ai, or check the dashboard if you have admin access.

My agent tried to edit config files and hit a security warning

Some AI agents ignore the one-liner above and try to hand-author ~/.claude/.mcp.json via cat > ... <<EOF, which triggers their own dotfile-overwrite scanner. Tell the agent: “Use the claude mcp add command exactly as shown — do not edit config files manually.”

What’s the tool access hierarchy?

Free, no key (5 tools): intel_search, intel_read, intel_feed, intel_knowledge, intel_briefing — all SQL-backed, cited, 10 req/min per IP.
Claimed key (+1 tool): intel_ask — synthesized LLM answers, 30 req/min per key, 90 days. Gated behind a key because it's the only tool that runs live inference; keys let us attribute cost + prevent abuse while keeping everything else free.
Master key (+1 tool): intel_ops — admin (ingest, source health, package management). Not exposed to public or package-scoped keys.