Documentation

Welcome to Kognita

Kognita turns your codebase and connected systems into a managed AI runtime. Developers get better context inside their AI tools, while management, product, support, and operations get direct access to grounded answers about how the system actually works.

What is Kognita?

Modern software teams spend hours each week answering the same questions: where is this function defined, which service owns this logic, what was the reason for this decision, what customer workflows might be affected by a change. Kognita indexes your repositories into a semantic layer that understands meaning — not just text — then runs a managed project agent on top of that layer so people can ask operational, product, and engineering questions without reconstructing the system from scratch.

That managed runtime gives non-technical teams immediate browser access to codebase-grounded answers, and exposes an MCP (Model Context Protocol) server for every project. Claude Code, Cursor, Codex, Windsurf, and other MCP-compatible assistants can use the same live system context while helping developers write, review, or debug.

What you get

Semantic code search

Find functions, patterns, and architectural decisions by meaning — not just keyword matches. Works across all indexed repos simultaneously.

MCP integrations

Connect Claude Code, Cursor, Codex, Windsurf, and any other MCP-compatible AI assistant with a single bearer token per project.

Custom MCP servers

Attach approved external MCP servers, including Zapier MCP, so the managed agent can use app actions and internal tools alongside codebase context.

Bring or host your own MCP

Point the always-on managed AI agent runtime at your own public HTTPS MCP endpoint using Streamable HTTP or SSE, with encrypted headers and project-specific guidance.

Multi-repo projects

Group any number of repositories under a single project. Index GitHub, GitLab, Bitbucket, or any public/private repo URL together.

Database MCP

Connect named read-only databases so approved teams can ask operational and customer questions in plain language.

Jira integration

Give your AI agent access to your Jira issues alongside your code. Reference tickets, understand requirements, and trace decisions.

How it works

Create an account, create an organization, then create a project for a service, product area, or multi-repo codebase.

Add a key, connect git, and add a repo

Save an Anthropic API key, connect GitHub, GitLab, or Bitbucket, and add at least one repository. That is enough to provision the project agent environment.

Provision and index

Kognita provisions the per-project agent runtime once the project has a key and a repo. Then index every repository in the project to build the semantic search layer.

Connect your AI tool via MCP

Open the MCPs section, copy your project's MCP URL and bearer token, and add them to Claude Code, Cursor, Codex, Windsurf, or any compatible assistant.

Anthropic API key

Kognita needs an Anthropic API key before it can provision a project's agent environment. Open your project in the dashboard, go to API Keys, paste your Anthropic key, and click Save.

Never shown again

After you save the key, Kognita only shows a masked version. The full key is never visible again in the dashboard.

Set a max spend

We suggest adding a max spend limit for the Anthropic key so you can cap usage and know exactly how much the project can spend.

Stored encrypted

Keys are encrypted before saving and used only to run the project agent environment tied to your project.

Replace anytime

To rotate the key, paste a new Anthropic key in the same API Keys screen and click Replace.

Who it's for

Developers — spend less time onboarding to unfamiliar code. Ask "how does authentication work?" and get a precise answer grounded in the actual codebase, not hallucinated summaries.

Engineering leads — give your team a shared, always-up-to-date understanding of your architecture without maintaining separate docs.

Non-technical teams — product, support, and operations can query what's in the codebase without filing tickets to ask engineers.

Next steps

Onboarding guide

Step-by-step walkthrough from account creation to your first indexed codebase.

Kognita MCP

What MCP is, how SSE and Streamable HTTP work, and how Kognita uses MCP in the managed runtime.

Semantic layer

Why indexing unlocks MCP context, reduces token waste, and grounds answers in actual code.

Git refresh

How Kognita keeps repository metadata, branch state, and indexed context current.

Jira integration

Connect your Jira account and give your AI agent access to your tickets.

Database MCP

Connect named databases so your managed agent can answer data questions in plain language.

System prompts

Save reusable instructions and rules that change how your managed AI agent answers in chat.

Custom MCP servers

Connect approved external MCP servers like Zapier so the managed agent can use app actions and internal tools.

Anthropic provider

Add an Anthropic API key, keep it encrypted, and set a max spend limit.

Keeping your environment safe

How Kognita secures your provisioned environment and how to set Anthropic spend limits and alerts.