Forage MCP: The Knowledge Layer Your Agent Is Missing

Riccardo Minniti / Ernesta Labs · director@useforage.xyz · useforage.xyz

Your agent hallucinates. Let's be honest about that.

You ask it: "Who are Stripe's main competitors?" It tells you confidently about companies that don't exist. You ask: "Find me 50 qualified leads." It returns hallucinations with made-up email addresses. You say: "What did you learn about this market last month?" It says: "I have no memory of last month."

You're not running an intelligent agent. You're running a smart model with amnesia.

Every research workflow starts from zero. Your agent searches, reads, answers, forgets. Tomorrow, when you ask about the same company, the same market, the same competitors—it learned nothing. Zero institutional knowledge. Zero compounding research.

Meanwhile, your competitor's agent? It's built a knowledge graph of 1 million entities. When it researches Stripe, it already knows the investors, the hires, the tech stack, the acquisition history. It adds what's new and moves on. Six months in, their agent knows your market better than you do.

What does it cost you when your agent is starting from zero every single time, while smarter agents are building permanent intelligence?

What if your agent actually had a memory?

Imagine your agent researched Stripe six weeks ago. Found funding, competitors, tech stack, hiring timeline, key executives. Today you ask about Stripe again. Your agent says: "I already know Stripe. Last time I found 15 competitors, Series D at $95B valuation, currently hiring 30 engineers on the infrastructure team. Here's what changed since then." No re-scraping. No hallucinations. Just: this is what we know, here's what's new.

That's not a research tool. That's an intelligence asset.

What would actually be different if your agent could:

• Run market analysis on Day 1 and still reference it on Day 180, watching hiring patterns, funding announcements, feature launches in real-time
• Answer "find me leads matching X criteria" and check its graph first—return cached results instantly, add new matches, never double-scrape
• Tell you "Company A and Company B share three investors" not from guessing, but from a 1-million-entity graph it built over months
• Track a prospect's hiring signals over time and tell you "they're moving from product to sales focus—here's when they'll be buying"
• Run causal analysis on market shocks and tell you which of your prospects will be affected and why

Step into that version of your workflow for a second. Every research compounds. Every discovery sticks. Your agent gets smarter, faster, cheaper every single day.

What does your competitive position look like then? What decisions become obvious that aren't obvious now?

The Knowledge Graph: What Compounds

Every tool call feeds a persistent knowledge graph. Your agent's research accumulates across sessions. The more it uses Forage, the smarter it gets.

Not a cache. Not a log. A structured graph with nearly 1 million entities and growing: companies, people, domains, funding rounds, competitors, technologies, markets, legal entities, geopolitical events, financial instruments. A FIBO-aligned ontology with 150+ entity types and 200+ relationship types including causal chains, contagion pathways, regime transitions, and shock propagation.

When your agent needs to know something, it checks the graph first. What it already knows, it doesn't re-learn. What is new, it adds. What is connected, it discovers.

This is the differentiator. Other tools give your agent a response. Forage gives your agent a memory that never forgets and a reasoning layer that connects everything it has ever learned.

36 Tools. One Connection.

Core Tools

... View full README on GitHub