Software & code
+125%signal lift · 2.25×
Saroir codebase · 1,672 nodes, 61 masters
Finds the handful of files that carry a large codebase, so agents and engineers focus where failure actually hurts.
Context engineering, perfected
Beyond context. Cognition.
Built to earn its keep.
Plugs into the agents you already use
ClaudeClaude CodeCursorContinueCodyZedWindsurfDevinopencodeAiderClaudeClaude CodeCursorContinueCodyZedWindsurfDevinopencodeAider
Get started
Sign in, or ask us anything.
Manage your plan, referrals and cloud projects, or chat with Webby about what Spiderbrain can do for your work.
Webby
Spiderbrain's support assistant
Hi, I'm Webby. What are you building, or what brought you to Spiderbrain today?
The shift
From context engineering to cognition.
Context engineering assembles what your AI knows. Spiderbrain takes the next step and understands it, reasoning about how everything connects the way you would. Same category, next rung.
The old wayContext Layer
Spiderbrain CortexCognitive Layer
User Profile
Agent Mind
Structured Files
Living Memory Graph
Context Distribution
Memory + Reasoning
MCP Access
Cognition Runtime
Context Management
Thought Management
Personal Knowledge
Persistent Intelligence
Validated on real data
It finds what matters. Proven on real data.
On 14 real-world datasets, from code to clinical trials, Spiderbrain pulled the few things that matter out of the noise. The bigger the signal lift on a card, the more sharply the important items stood out from the rest (we explain signal lift just below).
14 real-world datasets0 silent corruptions30 gates · 353 assertionsbit-for-bit deterministic
People analytics
+141%signal lift · 2.41×
Employee-attrition set · 1,470 records
Pulls the few factors that truly drive attrition out of hundreds of fields.
Insurance
+103%signal lift · 2.03×
Insurance risk set · 1,338 records
Ranks the high-leverage risk factors so underwriting attention lands where it counts.
Clinical trials
+48 to 60%signal lift · 1.48 to 1.60×
6 therapeutic areas · 6,000 trial records
Isolates the pivotal criteria across oncology, cardiology, diabetes, neurology and more.
Diagnostics
+91%signal lift · 1.91×
Breast-cancer set · 569 nodes
Catches hidden-critical features, including one that scored low alone yet ranked 4th most overlooked.
Geophysics & materials
+86%signal lift · 1.86×
Earthquakes · 2,000 nodes (largest tested)
Holds up on the biggest graphs we ran and still isolates the structurally critical signals.
Measured on real-world datasets (ClinicalTrials.gov, open ML sets, real codebases). Deterministic by design: same input, same result, every run. Read the full validation report →
What is “signal lift”?
Point Spiderbrain at a pile of things, files, records, trial criteria, and it scores how important each one is. Signal lift is how much higher the important few score than a typical item. Picture a 1,000-page manual where only a few pages are load-bearing: a signal lift of +140% (2.4×) means those pages score 140% more than the average page, so they jump out instead of hiding in the pile. The higher the lift, the easier it is to find the needle in the haystack.
Other metrics Spiderbrain generates
webscore
Severity. How much it would hurt if this one thing failed, on its own.Your login module scores 9.4/10. Lose it and everything goes down.
spikescore
Blast radius. If this breaks, how much breaks along with it.A shared config can be low-severity but high-spread: change it and dozens of things wobble.
blastVolume
Total risk in one number, combining severity and blast radius.The single “how worried should I be?” score for a dashboard.
Masters
Keystones. The few hubs everything leans on, which also act as firebreaks that stop damage spreading.One real codebase had 61 masters out of 1,672 files.
Blind-spot index
Important but unexamined. Things that score critical yet no one has reviewed or documented.Catches the file that quietly became load-bearing while no one was watching.
Drift
Going stale. Flags when something has changed in ways that no longer match its past.Tells you which knowledge has aged out, before it bites you.
How it works
Point it at your project. It builds a mind.
Three steps from a folder of files to an agent that genuinely understands your work.
01
Ingest
Aim Spiderbrain at any repo, folder or corpus. It parses every file, dependency, decision and change into one living memory graph that spans JavaScript, Python, Go, SQL, docs and more.
02
Reason
It scores every node by severity and blast radius, surfaces masters, blind spots and drift, and learns how failure spreads. The structural understanding flat context can’t give.
03
Connect
Wire it into Claude, Cursor, Continue or any MCP client. Your agents inherit persistent memory and reasoning that stays current automatically. Then forget it’s there.
$ spiderbrain import .
parsing 2,147 files across 8 languages
building memory graph ✓
scoring severity (webscore) ✓
computing blast radius (spike) ✓
ready (2,147 nodes, 9,612 edges, 14 masters)
$ spiderbrain why src/auth/session.ts
webscore 9.4 // severity
spikescore 8.7 // co-failure spread
blastVolume 17.6 // MASTER
if this fails → 42 dependents break
decision log → 6 signed entries
Capabilities
A mind, not a memory dump.
Everything Spiderbrain learns about your project, served to your agents as structured, queryable intelligence.
The context graph, perfected
Gartner expects most AI agents to run on a context graph by 2028. Spiderbrain turns that into a living memory graph: every file, dependency and decision as one connected, always-current model that reasons, not just stores. Explore it as an interactive living map and watch importance and blast radius light up as you explore.
How Spiderbrain stops context rot
LLMs lose accuracy as the window fills up. That’s context rot. Spiderbrain scores every node by importance and blast radius, so your agent gets the load-bearing few, not the noisy thousands.
1,000s of tokens→ the load-bearing few
Why Spiderbrain, not just RAG
RAGRetrieves similar text. No structure.
Other context toolsAssemble the context. Don’t reason over it.
SpiderbrainReasons over the context graph. Cognition.
Blast radius, answered
“What breaks if I change this?” gets a real answer. spikescore traces conditional co-failure across the graph, so risk is visible before you ship.
Severity that ranks itself
webscore rates every node by how much its failure would hurt, so your agent knows what actually matters.
auth/session9.4/10
api/router7.8/10
ui/button2.2/10
Cognition Runtime
A hosted MCP server. Connects Claude, Cursor, Continue, Cody, Zed and any MCP client. No glue code.
Thought management
Every decision is append-only, signed and hash-chained. A second brain you can audit: judgments, reasons and when they went stale.
Private & deterministic
Your source code never leaves your machine; only a source-free graph is scored. Same input, same brain, every time.
Reads everything your project is made of
Parses real imports across languages, plus documents and chat history, so the graph reflects the whole of your work, not just the code.
JavaScriptTypeScriptPythonGoJavaCSQLMarkdownPDFdocxChat history
Plans & pricing
Start free. Upgrade when it earns it.
Plenty of tools can build a graph of your project. Far fewer can tell you which parts are load-bearing. Spiderbrain scores every node by how much it matters and how far its failure would spread. That ranking is the moat, and the paid plans sharpen it.
Free
$0free forever
For small and medium projects.
- 1 project
- Native parsing on your machine
- Brain scored in the EU, map up to 1,500 nodes
- MCP access: Claude, Cursor, Continue and more
- Content-free usage counts you can switch off
- Free for personal, research and open-source use
No credit card required.
Spiderbrain Pro
$36$21per month
Founding members pay $21/mo forever
For as long as you keep the plan. Until Spiderbrain Cortex General Availability or while founding numbers last: 3,750 across the Pioneer, Founding Member and Genesis tiers.
Designed for people whose projects never stop moving.
- Up to 5 projects, map up to 3,300 nodes
- Cloud brain hosting
- Cross-device sync
- Support tickets through Webby
Upgrade from inside the app.
Cortex
$99per seat / month
For teams that share one brain.
- Everything in Pro
- One deterministic shared brain, map up to 9,000 nodes (plus 3,000 per seat)
- Predictable, auditable merging
- Assisted AI merge (beta)
- Bring your own model key
- On-call support
Sales-assisted onboarding.
Enterprise
Talk to us
For rollouts with specific needs.
- Everything in Cortex
- Commercial license
- Custom implementation by Perform Digital
- Priority support and a dedicated account manager
Custom licensing and rollout.
Failure spreads predictably. Spiderbrain proves it.
Deterministic graph science · webscore × spikescore · v5.1
Give your agent a mind.
Download Spiderbrain and build your first brain in minutes. Free for personal, research and open-source use.