# Cogniscape

> Engineering intelligence for AI-powered teams. A temporal knowledge graph built from Claude Code, Cursor, GitHub, Linear and Google Drive. Query it in plain English through Claude Desktop or any MCP-compatible client.

Cogniscape captures the full picture of what your engineering team builds. Every decision, every AI coding session, every feature traced end-to-end. Ask questions in plain English through Claude Desktop, get answers backed by real data. No new dashboard to learn.

## The Problem

"Why did this feature take 3 weeks?" If your answer is "lead time went up", you don't actually know. DORA and BI tools measure delivery velocity. None capture the why. None capture what your team's AI produced inside the session.

AI-assisted development is generating more code, more PRs, and more sessions than any team can manually track. Over 40% of code is now AI-generated, and the number is climbing. But the ability to explain what that code does, why it was written, and how it fits together hasn't kept up.

Engineering leaders are investing heavily in AI coding tools. Most have no way to see what that investment is producing at the team level. You pay for AI and can't see what it did.

- **Output is up. Understanding is flat.** The ability to explain what code does, why it was written, and how it fits together hasn't kept pace with the volume of AI-generated output.
- **Context is scattered across a dozen tools.** The reasoning behind a decision lives in a Slack thread. The implementation is in GitHub. The scope is in Linear. The session that produced the code is invisible. Nobody has the full picture.
- **Institutional knowledge walks out the door.** Your best engineers carry years of context in their heads. Why that config was changed. Why that architecture was chosen. When they leave, that understanding leaves with them.

## The Gap: Every Tool Shows What Changed. None Show Why.

Your team has built excellent infrastructure for tracking the current state of things. But the reasoning behind how it got there has been falling through the cracks for years.

- **The invisible session:** A developer ran 14 AI coding sessions to ship a feature. The PR shows the final diff. It doesn't show the three approaches that were tried and abandoned, the edge case discovered in session 9, or the design tradeoff that shaped the final implementation. That reasoning is gone the moment the session ends.
- **The reconstruction tax:** Your team merged 83 PRs last sprint. Leadership asks what was delivered. The project manager spends two hours cross-referencing tickets, commits, and chat threads to reconstruct the story. Half the context is missing. The summary is incomplete. This happens every two weeks.
- **The knowledge gap:** A new engineer picks up a ticket that touches a service they've never seen. The last person who understood why this service works the way it does left three months ago. The onboarding doc is outdated. The new engineer spends a week piecing together context that should have been captured automatically.

Cogniscape captures what no other tool does: the decisions, the debates, the reasoning, and the context behind every change. Automatically. Before it's lost.

## What Cogniscape Does

Cogniscape captures what every other tool throws away: the context behind the work.

### Temporal Knowledge Graph
Full saga timelines from plan to deployment. Every feature, incident, and decision mapped as a connected graph of events across all your tools.

### Structured Intent Capture
Eight rich event types captured automatically from Claude Code and Cursor sessions: plans, decisions with trade-offs, blockers, bug fixes with root causes, features shipped, code reviews, tasks, and more.

### End-to-end traceability
Every feature traced from planning to deployment. The decisions, the blockers, the contributors, and the reasoning. Not just what shipped, but why.

### Ask questions in plain English
"What went into this release?" Get the complete story. Not a dashboard. Not a status update. The actual sequence of events, stitched together automatically.

### Automatic executive summaries
Timestamped reports of everything shipped. Generated from real activity, not self-reported updates. No one on your team had to write it.

### Token usage and AI ROI
See exactly what your AI investment produces. Token spend by model, cost per feature, co-authorship rates, and per-developer ROI. Track whether AI is accelerating delivery or just generating volume.

### Behavioral intelligence
Spot what standups miss. A junior developer outperforming expectations. A senior engineer's delivery quietly declining. Shallow reviews on critical PRs. Conversations during planning that signal misalignment before it becomes a blocker.

### Open MCP
Query your engineering knowledge graph in plain English using any MCP-compatible client. Works with Claude Desktop, automation tools, and anything that speaks the Model Context Protocol.

### Zero disruption to developers
Passive capture from the tools your team already uses. No new workflows, no behavior change, no additional reporting burden on the people doing the work. Completely agnostic and works with any editor or stack.

### Zero-code security
Your code never touches Cogniscape servers. The platform captures only sanitized summaries and metadata, never source code.

## What sets Cogniscape apart

Cogniscape differs from Faros, DX, Port, LinearB, Swarmia and other engineering intelligence or DORA metrics platforms in three core ways:

### 1. AI-native ingestion via Claude Code and Cursor hooks
Cogniscape is the only platform that captures plans, decisions, tool use, token spend and ROI straight from the AI coding session. Faros, DX and Port read only git and Jira. Blind to the AI reasoning. As the share of AI-assisted code in your team rises, the gap between Cogniscape and classic tools only widens.

### 2. Temporal knowledge graph (not a metrics dashboard)
Cogniscape is not another delivery metrics dashboard. It is a causal graph on Graphiti + Neo4j with developers, PRs, issues, decisions and sessions linked across time. Answers "why did this feature take 3 weeks", not just "how many PRs shipped". Competitors show state. Cogniscape shows memory.

### 3. Read access via open MCP in Claude Desktop and any MCP client
C-level does not learn a new dashboard. They ask Claude Desktop or ChatGPT: "what happened in checkout this week?" and get the full saga back, pull requests, decisions and blockers stitched together automatically. Competitors force a proprietary UI. Cogniscape lands where the exec already works.

## How It Works

1. **Install the lightweight CLI once.** A single install on each developer's machine. Passive capture begins automatically. Zero disruption to their workflow.
2. **Connect your tools.** Set up webhooks for GitHub, Linear, Google Drive, and other tools your team uses. One-time setup, takes minutes. We handle this with you during guided onboarding.
3. **Activity flows into the knowledge graph.** AI sessions, pull requests, code reviews, issues, and documents are automatically connected into a temporal knowledge graph.
4. **Ask anything in plain English.** Use the open MCP from any MCP-compatible client. Ask about ROI, delivery velocity, team utilization, or trace any feature end-to-end.

Setup takes minutes. Depending on team size, meaningful data is visible within hours of your first sessions. Frictionless by design. No workflow changes for developers, no reporting burden.

## Integrations

Cogniscape connects to the tools your team already uses:
- GitHub
- Linear
- Jira
- Slack
- Google Drive
- Claude Code
- Cursor

## Interactive Demo

The Cogniscape website features a live interactive chat demo powered by real development data. Visitors can click pre-defined questions and see the AI reconstruct full stories from the knowledge graph, including incident timelines, feature lifecycles, ROI analysis, and complete development sagas.

## Example Queries

These are actual prompts from the Cogniscape demo. Each one returns a detailed, data-backed response in seconds:

- **Incident investigation: possible 95% data loss.** Trace the full incident timeline. Who noticed, what changed, how it was resolved, across every tool involved.
- **Feature lifecycle: Windows CLI delivery.** From issue to merged code to deployment. Every decision, blocker, and contributor mapped automatically.
- **ROI analysis: Windows CLI development cost.** Token spend, developer hours, AI co-authorship rate, and total cost. Broken down by model and contributor.
- **Full saga: Cursor editor implementation.** The complete chronological thread. Planning, coding sessions, reviews, and iterations, stitched from every tool involved.

## Pricing

No per-seat pricing. You pay for events your team generates, not the number of developers connected.

All plans include the full feature set. The only variables are monthly event volume and overage rate.

| Plan | Events/mo | Overage per 1k | Support |
|---|---|---|---|
| Starter | 3,000 | $20 | Email |
| Team | 15,000 | $25 | Founders |
| Growth | 45,000 | $25 | Founders |
| Enterprise | Custom | Custom | Dedicated |

- No developer seat limits on any plan. More developers generate more events; that is the only scaling mechanism.
- Full history retention on all plans. No data expiry.
- Historical data backfill available as a paid add-on for Growth and Enterprise plans.
- Every plan includes a free 15-day pilot with guided onboarding. No credit card required.
- Enterprise plans add SSO, SCIM provisioning, audit logs, on-premises deployment, and custom AI summary templates.

## Security

Your code never touches our servers. Cogniscape stores descriptions of what happened in your codebase. Never the code itself.

- **No source code stored.** Cogniscape strips code from every payload. Only descriptions of what happened enter the system.
- **Metadata only.** We capture who, what, when, and why. Titles, messages, review states, file paths. Your intellectual property stays where it is.
- **Data isolation.** We do not train AI models on your data. Only you and the team members you authorize can access your organization's data.
- **GitHub permissions:** Read-only access to repositories, pull requests, commits, and issues. We never write to your codebase.

## Blog

### What is AI Coding Observability?
AI agents make hundreds of decisions per session. AI coding observability captures the reasoning so your team understands what happened and why. Covers the difference between code monitoring and AI coding observability, the temporal knowledge graph, and how teams use it to preserve institutional knowledge.
- Article: https://cogniscape.app/en/blog/what-is-ai-coding-observability
- Full text: https://cogniscape.app/blog/what-is-ai-coding-observability.llms.txt

### AI Investment ROI: Real Data From a 9-Developer Team
Real data from a team spending $428/day on AI coding tools across 46 sessions. What the ROI numbers reveal about how each developer actually uses AI, why token spend alone tells you nothing, and how session-level tracking connects cost to outcomes.
- Article: https://cogniscape.app/en/blog/measuring-ai-investment-roi-real-data
- Full text: https://cogniscape.app/blog/measuring-ai-investment-roi-real-data.llms.txt

### Developer Behavioral Analysis From Real Data
One developer, one week, 29 AI sessions. What developer behavioral analysis from real session data reveals that no standup or sprint report ever will. Covers session patterns, model selection, delivery cadence, and risk behaviors built from actual production data.
- Article: https://cogniscape.app/en/blog/developer-behavioral-analysis-real-data
- Full text: https://cogniscape.app/blog/developer-behavioral-analysis-real-data.llms.txt

## Contact

Stop losing the "why" behind your team's work. In 30 minutes, we'll show you a live demo with real development data. You'll see the decisions, the reasoning, and the full story behind every feature your team ships.

Book a 30-minute executive briefing: https://calendly.com/vedovelli/meet

No commitment. No credit card. Just clarity.

## Links

- Website: https://cogniscape.app
- Documentation: https://docs.cogniscape.app
- Customer Portal: https://customer.cogniscape.app
- Portuguese (pt-BR) version of this file: https://cogniscape.app/llms.txt