Connect Snipara to the project you already work in.
The fastest path is folder-first: run the setup command from an existing repository or project folder, connect your AI client, and let Snipara keep the resulting Project Intelligence boundary in scope. Hosted MCP is the agent surface; API / SDK is for apps and integrators.
Start from the repository or local folder you already work in. You normally do not need an empty Snipara project first.
Run the guided command from the folder root, choose your AI client, and let Snipara create or match the remote project boundary.
Restart the client, ask it to run snipara_help, then ask a grounded question about the project.
Start from the folder your agent already knows.
Open a terminal at the root of your existing project and run npx create-snipara@latest. Choose your AI client, sign in or use an existing key, then restart the client. The folder name or Git remote supplies the project hint; you do not need to create a second empty project first.
npx create-snipara@latestGitHub and GitLab are optional. First value can come from the local README and project docs. Use snipara-companion only when local terminal workflows help.
- Snipara does not replace your LLM. It prepares the context you send to it.
- Hosted MCP is the canonical agent surface.
- API / SDK is the canonical app and integrator surface.
- `snipara_context_query` stays the default for docs and narrative knowledge.
- Memory stores durable decisions and preferences; it is not the document source of truth.
- Code structure lives in the code graph, not in a generic doc search result.
- Supported PDF, DOCX, PPTX, SVG, and VSDX files are parsed into searchable text during GitHub sync.
- Shared context is project-aware, so the same team can behave differently per project.
Four steps to the first grounded answer.
- Open your existing repository or project folder in a terminal.
- Run
npx create-snipara@latestornpx create-snipara@latest init --client <client>. - Complete browser sign-in or paste an existing Snipara API key.
- Restart the AI client, ask it to run
snipara_help, and ask one question grounded in the project.
Local folder and Snipara project are linked, not duplicated.
Your local folder is where you work. The Snipara project is its remote Project Intelligence boundary. create-snipara links them and reuses a matching project when one already exists.
No project folder yet? Use the dashboard's empty-project flow as a deliberate fallback, then upload your first document.
Diagnose the connection from the same project folder.
Tools missing
Restart the client or open a new session. Then run npx create-snipara@latest doctor --client <client>.
401 or permission error
Check that the API key and project slug belong together. Generate or reveal the key from the project's API key settings, then rerun setup.
No grounded answer
Let local indexing finish, ask snipara_help, and retry with a question about the README or project docs.
Still blocked? Use the complete create-snipara troubleshooting guide.
Code Graph v1
In v1, snipara_context_query stays doc-first. It remains the right default for docs, notes, markdown, and narrative context.
For structural code questions, use snipara_code_callers, snipara_code_imports, snipara_code_neighbors, or snipara_code_shortest_path.
HTTP direct is the cleanest hosted setup.
If your client accepts remote MCP, use the project slug and an API key. Keep the slug as the human-readable identifier across all transports.
{
"mcpServers": {
"snipara": {
"type": "url",
"url": "https://api.snipara.com/mcp/YOUR_PROJECT_SLUG",
"headers": {
"X-API-Key": "snp-YOUR-API-KEY"
}
}
}
}The example uses the remote-MCP type/url shape. Some clients use a different configuration key or transport label; follow the client-specific integration page when its schema differs.
Integration guide
Compare hosted HTTP, SSE, and local stdio without changing the story every time.
Shared context
Learn how reviewed memory and project-scoped context travel between runs.