Onboarding

Start from the Kognita sign-up screen. Create an account with your email and password, then sign in to reach the dashboard.

The account is your user identity. Organizations, projects, repositories, API keys, and provisioned agent environments are created after you sign in.

An organization is the top-level workspace that holds your projects and git provider connections. You'll need at least one before you can do anything else.

In the dashboard sidebar, click + New organization. Enter a name (usually your company or team name) and optionally your website URL. Click Create.

You can have multiple organizations — useful if you work across separate companies or need to keep teams completely isolated.

Projects hold repositories, API keys, indexing runs, MCP credentials, and the optional per-project agent environment. Think of a project as a named codebase, product area, or service boundary.

From the Projects view, click + New project. Give it a name — typically the name of your service, repo, or product area. Click Create.

After creating, a modal will show your project's MCP URL and bearer token. You can always retrieve these later from the MCPs section of the sidebar.

The agent environment needs an Anthropic API key before it can be provisioned. OpenAPI keys for the project and save your Anthropic key.

After the key is saved, the full value is never shown again. Kognita stores it encrypted and only displays a masked version so you can confirm that a key exists without exposing the secret.

We recommend setting a max spend limit for this key in Anthropic. That gives you a clear ceiling for usage and makes it easier to understand how much the project is spending.

If you add the key before adding a repository, Kognita will provision the environment after the first repo is added. If you add the repo first, provisioning starts after the Anthropic key is saved. Either order works.

Before you can add repositories, you need to authorize Kognita with your GitHub, GitLab, or Bitbucket account. This connection lives at the organization level and can be shared across all projects in that org.

Go to Git connections in the sidebar and click Connect next to your provider. You'll be redirected to that platform to authorize access, then returned to the dashboard with the connection active.

With a git provider connected, you can now add repos to your project. Open the project from the sidebar or projects list, then click + Add repo.

A modal will appear showing your connected providers. Click the provider whose repo you want to add — it becomes selectable only if the connection is active. A searchable dropdown will list all repos accessible through that connection. Select the repo you want, confirm or change the branch (defaults to the repo's default branch), then click Add.

Repeat this for as many repositories as you need in the project. You can mix repos from different providers within one project.

Added repos start with a not indexed status. Indexing is what builds the semantic layer that makes your code searchable.

In the project view, click the Index all button. Kognita indexes every repository in the project together, so indexing is an all-or-none project action. A confirmation modal will show the repositories that are about to be indexed.

Status changes to indexing… while processing, then to indexed when complete. Indexing typically takes 5–30 minutes depending on codebase size.

Provisioning creates the project's dedicated agent runtime. This is the environment Kognita uses for project-aware chat and connected MCP tools.

The requirement is intentionally light: the project needs an Anthropic API key and at least one repository. The repository does not need to be indexed before provisioning can start.

In the chat panel, click Provision agent environment if it appears. In the normal onboarding path, Kognita also attempts provisioning automatically after the project has both prerequisites. Setup usually takes about a minute.

If you use a local code assistant — Claude Code, Cursor, Windsurf, Codex, or any MCP-compatible tool — you can point it at your project's Kognita MCP endpoint so it gets the same semantic search context as the cloud agent.

Open MCPs in the sidebar and select your project. You'll see two credentials:

• URL — the MCP server endpoint for this project
• Authorization — the bearer token that authenticates your assistant

Copy both and add them to your assistant. The exact configuration depends on the tool:

The MCPs tab is also where you wire up the rest of your agent's context: add custom MCP servers, database connections, a Jira connection, and other integrations. Anything you connect here becomes available to your AI assistant alongside the semantic code search.