AI-native sprint memory for engineering teams

Your team's decisions
deserve to be remembered.

Cortex extracts organizational knowledge from your daily work — chat updates, meetings, PRDs, blockers — and builds a living knowledge graph your whole team can query. No extra process. No new rituals.

Used by teams spending 4+ hours a week re-discovering past decisions.

scroll
The silent knowledge drain

You're not losing code. You're losing context.

The tools you use remember what was built. Cortex remembers why.

🔁

"Why was this built this way?"

Every sprint ends, engineers move on, and the reasoning behind decisions lives only in Slack threads and the memory of the person who wrote the ticket — if they're still around.

🗺

"Which team owns this system?"

New hires spend weeks reverse-engineering relationships between teams, domains, and systems that senior engineers know by instinct. That knowledge is nowhere structured.

🔄

"We already discussed this."

Action items from planning meetings live in notes no one revisits. Two weeks later, the same question comes up again in standup.

A memory layer that builds itself.

Cortex runs alongside your existing tools. As your team sends chat updates, imports sprint plans, uploads PRDs, and records meetings, it continuously extracts and organizes durable knowledge — automatically.

Capture

Works from what you already do

Chat updates, DOCX sprint plans, PDFs, and meeting recordings. Cortex ingests from the channels your team already uses — no new forms to fill out.

Organize

Structured, queryable, connected

Knowledge is classified, deduplicated, and linked into a live graph. Rules, constraints, capabilities, team ownership, dependencies — every node traceable back to the conversation that created it.

Retrieve

Ask in plain language

Chat the question. The engine searches across tasks, FAQs, knowledge events, and meeting summaries — surfacing the answer with its source.

Four memory systems. One coherent picture.

Core memory

Organizational Knowledge Base

Every rule, constraint, and capability — extracted from natural conversation.

As your team discusses sprint work, Cortex listens for durable knowledge: team constraints, architectural rules, dependency boundaries, risk signals. It ignores the transient and captures the reusable.

  • Types: constraints, capabilities, dependencies, risks, process rules
  • Scopes: team · system · process · project
  • Semantic similarity links via vector embeddings
  • Manual override with /force-kb: when you need authority
  • Version-chained — see when a rule changed and what replaced it
Org topology

Entity Relationship Graph

A live map of who owns what.

Cortex maintains a canonical entity store — every team, system, domain, person, process, and constraint your work touches. Each entity is extracted, normalized, deduplicated by name similarity, and connected via typed relationships.

  • 7 entity types · 8 relation types
  • Alias merging: "Payments" and "Payments Team" → same node
  • Hierarchy resolution: parent/child chains built automatically
  • Full provenance: every node traceable to its source conversation
The full picture

Unified Cross-Source Graph

Tasks, decisions, meetings, and docs. One graph.

The unified graph pulls from all five stores simultaneously — KB facts, sprint tasks, task history, FAQs, and meeting summaries — and builds semantic edges between them based on meaning, not just metadata.

  • 5 source types: KB facts, tasks, task events, FAQs, meetings
  • Semantic edges: cosine similarity over independent embeddings
  • Structural edges: task_event → task (owns)
  • Updates on every chat message via background synthesis
Task history

Task Memory & Digests

Every task remembers its full story.

Each task accumulates an append-only event log from the moment it is first mentioned. Cortex condenses that log into a rolling digest — a short LLM narrative of the task's life so far.

  • Event log: deduped by content hash — no noise from repeated updates
  • Digest: rebuilt only when new events arrive (cost-efficient)
  • Timeline query: works in plain chat
  • Backfill: retroactively reconstruct history from existing logs

Up and running in one sprint.

No migration. No workflow changes. Cortex layers over your existing rhythm.

01

Import your plan

Paste your sprint plan or drop a DOCX. Cortex parses the task table, initializes state, and immediately syncs task definitions into the knowledge base. No migration, no data mapping.

# DOCX import
Import → paste path to your sprint plan
# Or start from Markdown
workspace/sprint_plan.md
02

Chat your updates

Update tasks the same way you'd tell a teammate. "Billing integration is blocked — Stripe needs a new webhook config before we can proceed." Cortex identifies the task, updates its status, extracts the constraint, and logs everything.

> Billing integration is blocked —
Stripe needs a new webhook config
✓ Task updated: blocked
✓ Constraint captured: kb_events
✓ Daily log appended
03

Query anything

"Which tasks are at risk this week?" "What does the auth team own?" "Show me the history of the mobile push notification task." Plain language. Instant answers with full provenance.

> What does the auth team own?
Auth Team owns:
→ Auth System (system)
→ Session Rules (constraint)
→ Onboarding Flow (process)
Source: 3 chat messages, 1 PRD
Integrations

Your knowledge, understood by every AI.

Cortex exports clean, structured context — paste it straight into any system prompt or use our native integrations with the models and tools your team already relies on.

Language models
OpenAI
GPT-4o
Claude
Claude 3.5
Gemini
Gemini 1.5 Pro
Mistral
Mistral Large
Llama
Llama 3.1 405B
DeepSeek
DeepSeek R1
Perplexity
Sonar Pro
Ollama
Local models
HuggingFace
Inference API
OpenAI
GPT-4o
Claude
Claude 3.5
Gemini
Gemini 1.5 Pro
Mistral
Mistral Large
Llama
Llama 3.1 405B
DeepSeek
DeepSeek R1
Perplexity
Sonar Pro
Ollama
Local models
HuggingFace
Inference API
Dev tools & IDEs
Cursor
AI Code Editor
Windsurf
Agentic IDE
GitHub Copilot
Workspace
Zed
Collaborative IDE
Notion
Connected docs
Linear
Issue tracking
Obsidian
Knowledge base
Raycast
AI commands
Cursor
AI Code Editor
Windsurf
Agentic IDE
GitHub Copilot
Workspace
Zed
Collaborative IDE
Notion
Connected docs
Linear
Issue tracking
Obsidian
Knowledge base
Raycast
AI commands

Simple pricing. Scales with your team.

MonthlyAnnual save 20%

Starter

For small teams getting their sprint rhythm right.

$49/mo
Start free trial
  • Up to 5 users
  • 1 active sprint
  • Chat updates & task tracking
  • Daily logs & status tables
  • Gantt view
  • Overall KB (up to 500 events)
  • Task history
  • Meeting transcription
Most popular

Team

For teams where institutional knowledge is a competitive edge.

$149/mo
Start free trial
  • Up to 25 users
  • Unlimited sprints
  • Everything in Starter
  • Entity-relationship graph
  • Unified cross-source graph
  • PRD ingest (up to 20 PDFs)
  • Sprint report generator
  • FAQ system

Scale

For orgs where knowledge loss is a documented cost.

$399/mo
Talk to us
  • Unlimited users
  • Unlimited sprints
  • Everything in Team
  • Semantic threshold tuning
  • API access
  • Priority support + onboarding
  • SSO (SAML)
  • Graph refinement controls

All plans include a 14-day free trial. No credit card required. Your data stays in your cloud tenancy — we don't train on your sprint content.

Common questions