Deterministic context layer for AI

Your AI forgets. Your business can’t.

Businesses don’t need more AI. They need executable intelligence.

As AI agents multiply, so do conflicting answers, forgotten context, and inconsistent decisions. A deterministic context layer gives every system one governed understanding of your business, so AI is consistent, explainable and ready to act. These are answers with receipts: every brain answer arrives with a receipt you can re-run.

Start building a brainSee the validation
Plugs into the agents you already use
ClaudeClaude CodeCursorCopilotCodexContinueCodyZedJetBrainsWindsurfAntigravityKiroDevinopencodeAider& others
Finance
People
Engineering
CRM
ERP
Documents
SpiderbrainDeterministic brain
Reports
Agents
Apps
Dashboards
Governance
Compliance
APIs
Insights
One Brain. Every conversation.

Why a business needs a deterministic context layer

Spiderbrain is a combination of multiple layers that stack together, and every layer above depends on the one beneath it holding true. When these layers work together you get reproducible answers, shared team understanding, pre-change risk visibility, and governance that actually sticks. Without them, AI remains clever but ultimately clueless about the real structure and real-time changes of your business. If you are already weighing this against a semantic layer or a data catalog, here is how a context layer differs from a semantic layer.

  1. 01Execution layer

    Agents and tools act on real structure instead of guessing, all against the same ground truth.

  2. 02Context layer

    A live map of what depends on what, calculated from source, so impact is visible before it ships.

  3. 03Semantic layer

    A shared dictionary that turns fragmented data into business meaning an agent can act on.

  4. 04Memory layer

    Persistent, versioned and shared, so the understanding survives the session and the team.

  5. 05Compliance layer

    Policies and regulations live in the brain as typed, versioned artifacts, not prompt suggestions.

  6. 06Governance layer

    Nothing governs an agent until a person ratifies it, and every rule carries who approved it.

E-comm use case: how Spiderbrain helps businesses grow

A customer buys running shoes. The shoes are too big. The customer asks to swap them for a smaller size. Three agents handle the rest.

Three agents, three memories

no shared brain
  1. 1Support agent

    Accepts the return. Writes it down in its own notes.

  2. 2Billing agent

    Does not know it is an exchange. Sends a full refund, then asks the customer to pay again.

  3. 3Sales agent

    Does not know anything changed. Keeps showing ads for the same shoes, in the size the customer just sent back.

What it costs you

The customer hears three different stories from one company. Your ad money is spent selling the size they returned. They ask for a full refund instead, and they do not come back.

Three agents, one Spiderbrain

one shared brain
  1. 1Support agent

    Accepts the return and records it in the shared brain.

  2. 2Billing agent

    Reads the brain. Sees an exchange, not a refund. Keeps the payment and charges only the small difference.

  3. 3Sales agent

    Reads the brain. Stops those ads and shows the smaller size instead.

What you keep

One company, one answer. No ad money wasted. The customer gets the right shoes in a few days, keeps the order, and buys again.

The artifact

One brain, every question answered the same way

The brain can plug into all your data sources, keep itself updated, collect real-time signals, and help agents make lightning-fast decisions. It is persistent, versioned, auditable and reusable. Connect any AI through MCP, embed it with the Spiderbrain API, or build new products on a foundation every team can trust.

source-freeversionedcontent-addressedprivate by default
graph / dependency map
masterkeystonedrift
Spiderbrain Explained

Every business will adopt AI. The businesses that win will build a robust Brain first.

Local machine

Private data and code

Privacy boundarySource stays. Understanding crosses.
MCP / API
Driftstale knowledge
Blind-spotunseen load-bearing
Masterskeystones
Webscoreseverity
BlastVolumetotal risk
Spikescoreblast radius
Compliance reportsgovernance
AI agentconnected via MCP
Business insightsdata strategy
SBMemory tree10 activev28
Spiderbrain memory tree: the decisions, constraints and patterns the brain remembers, each one ratified.
Governed memory
SBVisualize · blast-riskwebscore 9.6blast 49
Spiderbrain graph view colouring every file by blast risk, with the selected file showing its scores and dependents.
Scored file web
  1. 01

    Spiderbrain analyzes your files locally and builds a deterministic, governed brain that represents your organization’s understanding.

  2. 02

    Unlike AI models, the brain is persistent, versioned, auditable, and reusable.

  3. 03

    Connect any AI through MCP, embed it into your own applications with the Spiderbrain API, or build entirely new products on a foundation that every team can trust.

Spiderbrain API

One brain, plugged into everything that needs to know

The Spiderbrain API lets software ask questions about the relationships, decisions, constraints, history, and consequences inside a system. That makes Spiderbrain useful anywhere there is a complex system that changes over time.

  • Inside your own apps

    The moment something breaks, your backend can ask how far the breakage reaches, before a customer finds out.

    • Impact analysis before deployments
    • Dependency analysis
    • Architecture drift detection
    • API contract checking
    • Change-risk scoring
    • Regression investigation
    • Incident root-cause exploration
    • Understanding legacy systems
    • Migration planning

  • Docs and knowledge bases

    Ship a new product or land new research, and the pages it just made stale get flagged instead of quietly rotting.

    • Detecting stale documentation
    • Undocumented architectural decisions
    • Broken citations and references
    • Pages affected by new research
    • Contradictory or duplicate knowledge
    • Decisions with no recorded rationale

  • Auditor portals

    Run it before every audit call, so auditors open the current picture rather than an export from last quarter.

    • Exactly what changed since the last audit
    • Reproduce the understanding as of the audit date
    • Which evidence supports this control
    • Which controls a regulation change touches
    • Requirements with no evidence attached
    • Approvals and who granted them

  • Security reviews

    See where risk actually sits, and how far it spreads if that piece moves.

    • What depends on this vulnerable package
    • How far the exposure propagates
    • Which customers are exposed
    • Which controls mitigate the risk
    • What changed since the last security review
    • Known relationship versus model inference

  • Sales enablement

    Reps see the current offers and what each one does to a product bundle, before they promise it.

    • Which products are compatible
    • Which discounts conflict
    • Which customer segments are excluded
    • Which contract terms supersede standard pricing
    • Which product dependencies make an offer invalid
    • Which claims have no current evidence

  • Analytics

    Point queries at the blind spots nobody thought to ask about.

    • Which reports depend on this upstream system
    • Where a metric is defined more than once
    • Which blind spots carry the most load
    • What changed since this dashboard was built
    • Which conclusions depend on this dataset
    • Relationships no dashboard is tracking

How do you plan to use the Spiderbrain?Read the API docs

How it works

From repository to consultable ontology in minutes

01

Parse

Point Spiderbrain at any folder or repo. The native parser reads real imports on your own machine and builds a source-free dependency graph. Your file contents stay put; only the derived structure travels.

02

Score

Deterministic scoring finds keystones, blast radius, blind spots and co-failure paths. Same input, byte-identical output, every time.

03

Release

Publish a brain release: an immutable, content-addressed snapshot with provenance. Restore any one, and diff two to see what your understanding gained or lost.

04

Consult

Read it yourself in the app, in reports and in pull requests. Or connect Claude, Cursor, Continue or any MCP client, and every agent inherits the graph, the scores and your ratified decisions in every session.

Trust, in public

Do not take determinism on faith. Audit it

Determinism is only worth anything if you can check it. The methodology, the results and the gaps are all published, including the time we found a defect in our own engine.

  • Open reproducibility benchmark

    contextbenchmark.com publishes our methodology, machine-readable results and trust levels. When we found a defect in our own engine, we published the defect, the fix and the verified re-run the same day.

    contextbenchmark.com (opens in a new tab)

  • Signed transparency log

    Every brain in the public registry is recorded in a signed, append-only log. What was published, when, and with which fingerprint.

    View the log

  • Runs where you need it

    Parsing is native and runs on your own machine. Scoring and brain hosting run on our EU infrastructure today. Because only the derived understanding travels, never your file contents, where the brain lives is a choice rather than a risk: talk to us about regional or self-hosted deployment.

    Read the data governance policy

  • det-score

    The brain tells you what it does not know. Coverage is disclosed per repository, never silently overstated.

    How det-score works

Measured on 14 real datasets, from code to clinical trials, with every figure and the engine version it was measured on recorded against it.

The comparison

Three ways to give an AI context. Only one of them can prove it

An agent on its own guesses from similarity. A memory tool remembers what it was told, extracted by a model. Spiderbrain derives the structure from your source and can show its work. The job being compared is keeping an AI reliable about the systems you already run.

What you needAI agent aloneAI + memory toolAI + Spiderbrain
Trust
Same input, same context, every timenonoyes
Built from your source, not inferred by a modelnonoyes
Answers you can re-run and verifynonoyes
Tells you what it does not knownonoyes
Understanding
Knows what depends on whatpartialnoyes
Knows what breaks before you ship itnonoyes
Useful even with no AI in the loopnonoyes
Governance and time
Survives the sessionnoyesyes
Shared by the team, not rebuilt per seatnoyesyes
Versioned, so you can diff what changednopartialyes
Your file contents never have to leave your machinepartialnoyes

Determinism is a property of the map Spiderbrain builds, not of the sentences a model writes from it. The receipt is what makes an individual answer checkable. Characterisations of other categories reflect public product information as of July 2026 and describe categories, not named products.

What the brain measures

Understanding Starts With Measurement

Every file gets scored, so the load-bearing few surface on their own instead of drowning in a context window. Same input, byte-identical output.

webscore
Severity
spikescore
Blast radius
blastVolume
Total risk
masters
Keystones
drift
Stale knowledge
blind-spot
Unseen load-bearing
det-score
Derived, not guessed
coverage
How much it admits
clusterspikescore
Damage kept local
Guidelines

Your rules, ratified. Your agents, governed

Bring your policies and architecture documents. Each rule becomes a typed directive that waits in quarantine until a person ratifies it. Only then does it govern the agents that consult the brain, and it carries who approved it and when. Retire a rule and it stops applying, but the audit trail stays.

Download the EU AI Act packearly access
directive / quarantine
Distilled from · EU AI Act, Article 14

High-risk AI output requires effective human oversight. Agent-proposed changes to deployment-gating code must be reviewed by a person before merge.

pendingquarantine
Transparent plans and pricing

Start Free. Scale With Confidence.

Signing in gives you $5 in API credits so you can start building today. Use our API and MCP to create your own solution, or choose a managed plan and let our team take care of the setup, deployment, and scaling.

SpiderbrainThe brain. For developers and teams.

Additional tier benefits apply

Conductor SMEIntro priceThe platform. Your business data on the brain.

Spiderbrain engine included, as API access not the desktop app

Conductor OrgShared brain, agentic CDP, governance.

Spiderbrain engine included, as API access not the desktop app

User typeImplementationData LocationData ingestionParserData storageBrain storageSync brain to GitHubContext LayerMemory LayerNumber of projectsPrivacy & receiptsIdentity resolutionAgentic CDPSemantic LayerData segmentationAccess to 100+ agentsExecution LayerCDP profilesEvent volumeManaged setupCompliance Evidence PackDeployment
Managed Set-up (SME)
Choose a ring to see just that layer
01 BASIC SETUP & PARSING
  • User typeTeam
  • ImplementationManaged
  • Data LocationMulti-source
  • Data ingestionAuto API
  • ParserLocal + Cloud
  • Data storageDistributed
  • Number of projectsNot metered
  • Privacy & receiptsIncluded
02 CORE ARCHITECTURE
  • Brain storageDistributed
  • Sync brain to GitHubIncluded
  • Identity resolutionIncluded
  • Semantic LayerIncluded
  • Context LayerIncluded
  • Memory LayerIncluded
03 INTELLIGENCE & AGENTS
  • Data segmentationIncluded
  • Access to 100+ agentsIncluded
  • Execution LayerIncluded
  • Agentic CDPIncluded
  • CDP profiles50,000
  • Event volume5M / mo
  • Shared brain in teamnot included
04 ENTERPRISE & GOVERNANCE
  • AI Assisted mergenot included
  • Brain as a servicenot included
  • Commercial licensenot included
  • Managed setupIncluded
  • Compliance Evidence PackAdd-on
  • Governance Layernot included
  • DeploymentSaaS
Generous tier benefitsSpiderbrain

Spiderbrain Pro

$36$21per month

Founding members pay $21/mo forever

For as long as you keep the plan. Until 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
  • The brain only. Conductor platform sold separately.
  • Includes $20 of API usage every month

Secure checkout. Cancel anytime.

Intro priceConductor

Self Set-up

$999$498per month

+ one-time setup, priced by region

For a business putting its own data to work.

  • Your own single-tenant instance, on your own domain
  • Spiderbrain engine included as API access, not the desktop app
  • Agentic CDP: up to 25,000 profiles
  • We provision and stand up the instance
  • Your tech team configures data collection, rules and identity
  • Multi-source ingestion, 2.5M events a month
  • Identity resolution across your sources
  • Semantic, context and memory layers
  • Data segmentation
  • Access to 100+ agents
  • Execution layer, scheduled by Webby
  • Compliance evidence pack (add-on)
  • Includes $99 of API usage every month

We provision it. Your team configures it.

Special launch offerIntro priceConductor

Managed Set-up

$1,500$750per month

+ one-time setup, priced by region

~5 consulting hours a month

The same platform, set up and run for you.

  • Everything in Self Set-up, still single-tenant
  • Agentic CDP: up to 50,000 profiles, 5M events a month
  • We run the set-up end to end: collection, rules, identity, configuration
  • Connector work and tuning included
  • Priority onboarding
  • Includes $270 of API usage every month

We provision it and configure it.

Conductor

Cortex

$3,000-5,000per month

+ $99 per seat, plus one-time setup

~10 consulting hours a month

For an organisation running on one brain.

  • Everything in Managed Set-up
  • On-premise installation available
  • Agentic CDP: up to 250,000 profiles, 25M events a month
  • 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
  • Brain-as-a-Service API (REST + MCP)
  • Governance layer (add-on)
  • On-call support

Assisted onboarding.

Conductor

Command

Talk to us

+ one-time setup, priced by region

~45 consulting hours a month

The driving force behind an enterprise that intends to lead its industry and stay there.

  • Everything in Cortex
  • Full Brain-as-a-Service API access
  • Commercial license
  • On-premise or your own VPC
  • Profile and event volume set to your rollout
  • Compliance evidence pack included
  • Custom implementation by Perform Digital
  • Priority support and a dedicated account manager

Custom licensing and rollout.

Deployment

Runs where your data has to stay.

Because only the derived understanding ever travels, never your file contents, where the brain runs is a choice rather than a risk. Which of these is open to you depends on the plan.

01 / On premises

In your own data centre.

Conductor is installed inside your own perimeter, on infrastructure you control, and provisioned for your organisation alone. Available on Cortex and Command.

02 / Your own cloud

In your own VPC.

Conductor runs in a cloud account that belongs to you, on a dedicated instance sized to your scale, under your own domain. Available on Command.

03 / Local

On your own machine.

The Spiderbrain desktop app parses natively on your laptop. Your source code, file contents, commit messages and author emails never leave it. Only a source-free graph travels, for scoring.

How Conductor is deployed per client · What travels and what does not

FAQ

Questions, answered

01 What is the concept of the second brain? +
Remember Dumbledore’s Pensieve from Harry Potter? A magical stone basin used to store, review, and dive into memories? That’s essentially what a second brain (also Spiderbrain) is. Somewhere to keep what you know outside your head, so you and your tools can use it later. People have built them for years out of notes and wikis, and they go stale: someone has to write the page, remember to update it, and notice when it stops being true. Mostly nobody does, and an AI reading it will confidently give you last year’s answer. Spiderbrain is a second brain that maintains itself. It reads your actual systems, works out how they connect, and re-reads only what changed, so the understanding stays in sync without anyone remembering to update it. It is also versioned, so you can go back to any earlier state; governed, so nothing becomes fact until a person approves it; and shared, so every team and every AI reads the same thing.
02 What does deterministic actually mean here? +
The same repository produces the same brain on every machine and every run. No sampling, no drift. That is what lets an agent trust the answer, and what lets you re-run it.
03 Does my code leave my machine? +
Local scans never persist your source. Only a source-free graph is scored. The cloud path is opt-in and records provenance, so you always know what was read.
04 Which AIs can use the brain? +
Any MCP-capable host, including Claude, Claude Code, Cursor, Continue, Zed and Windsurf. Point them at your brain and they answer grounded in it.
05 Is it useful without an AI in the loop? +
Yes. The scores, keystones and blast radius are readable in the app, in reports and in pull requests, with no agent involved.

Getting started is easy

Build your first brain in under a minute and turn what your codebase already knows into executable intelligence. Your source stays on your machine.

Get started freeTalk to us