Klarix Intelligence Engine MCP¶
Remote MCP server (Streamable HTTP, keyless) for B2B competitive and sales intelligence built from cited public-web evidence. Fifteen tools, live-verified: semantic prospect matching over tens of thousands of pre-profiled companies, five-axis ICP fit scoring, live competitor battlecards, head-to-head comparisons, buying-committee mapping, account trigger detection, and institutional-grade tech-stack and product teardowns. Every answer is grounded in public web sources the agent can cite.
Server type: Remote (Streamable HTTP)
Auth: None required for reads (keyless)
Endpoint: https://mcp.klarix.ai/mcp
Tools: 15 (prospect matching, ICP scoring, battlecards, research, teardowns)
Pricing: Hosted; docs at klarix.ai/docs/api
Category: Sales & Competitive Intelligence
Built by: SpeaksenseAI (github.com/SpeaksenseAI/klarix)
Why This Matters for Operators¶
The pre-call homework that separates a winning sales motion from a cold one - who the buying committee is, what changed at the account, where the incumbent is weak - usually lives in three tools and a consultant. Klarix compresses it into one endpoint the agent can call mid-conversation. The ICP scoring is the operator-friendly part: score a company against YOUR ideal customer profile on five weighted axes (Strategic Fit, Tech Alignment, Buyer Persona Readiness, Growth signals, and more) instead of renting a fit model. The battlecard and teardown tools run live research, so the SWOT and the displacement playbook carry current, citable evidence rather than last quarter's deck.
Live-verified keyless. The endpoint answered an anonymous MCP probe (server klarix-intelligence v0.1.0, protocol 2025-03-26) with all 15 tool names - no account needed to enumerate the surface.
Tools & Capabilities¶
| Tool | What it does |
|---|---|
find_matched_prospects |
Semantic search over pre-profiled B2B companies (Voyage 4 embeddings, sub-100ms) |
score_prospect_fit |
Five-axis ICP fit scoring per account |
get_competitor_battlecard |
Live SWOT and positioning gap for a competitor domain |
analyze_head_to_head |
Your company vs a specific competitor, researched both sides |
get_company_intelligence |
Executive one-pager from live public sources |
generate_outreach_sequence |
Cold outreach that opens on a researched, dated insight |
get_market_landscape |
Tier-1 incumbents vs emerging challengers for a vertical |
get_buying_committee |
Economic Buyer, Champion, Technical Evaluator, Procurement mapping |
detect_account_triggers |
Leadership changes, funding, hiring surges, product updates |
teardown_tech_stack |
Technical stack deconstruction with displacement opportunities |
generate_displacement_playbook |
Switching-cost mitigations, POC blueprints, migration plans |
teardown_product_spec |
Hardware, battery, medical or API teardown to institutional grade |
Plus generate_swot_analysis, get_deep_research and search_scientific_evidence for citable literature.
Installation¶
claude mcp add --transport http klarix https://mcp.klarix.ai/mcp
No key. The same endpoint works in Claude Desktop, Cursor, VS Code and any Streamable HTTP MCP client.
Configuration¶
None required. For higher-rate or gated tool access, see the API docs at klarix.ai/docs/api.
Business Relevance¶
Founders and revenue operators get a standing competitive-intelligence function inside their agent: score inbound leads against the ICP before the SDR touches them, generate battlecards on demand for pipeline reviews, and teardown a competitor's stack before a positioning conversation. The cited-evidence design means the outputs survive a skeptical review.
Integration with CorpusIQ¶
CorpusIQ connects your own business systems (CRM, ads, commerce, finance) as read-only context. Klarix adds the external layer: prospect fit, competitor posture and market landscape. Together they give an agent both sides of a sales or expansion decision - internal performance from CorpusIQ, external evidence from Klarix - in one conversation.
Limitations¶
- Pre-profiled company universe is tens of thousands, not the full B2B web; long-tail accounts may miss the vector index
- Early-stage (v0.1.0) - tool surface and pricing may shift
- Anonymous probe succeeded, but high-volume commercial use may require a key