Configurations overview
A configuration is a named set of connection details, saved in your project and reusable everywhere in it. One Postgres connection, one Anthropic API key, one Salesforce login — created once, referenced by name.
Nothing that uses a configuration ever contains its values. A tool stores {{config.postgresql.orders_db.password}}; the password itself stays in the project, encrypted, and is resolved at run time. Rotate the key in one place and everything picks it up.
What you can configure
The category tabs across the top group configurations by what they connect to.
| Category | What goes here | Examples |
|---|---|---|
| Models | LLM providers | OpenAI, Anthropic Claude, Azure OpenAI, Google Gemini, Google Vertex AI, Vertex AI Claude, AWS Bedrock, Comet API, Ollama |
| Embeddings | Embedding models for search and RAG | OpenAI, Azure OpenAI, Cohere, Voyage AI, HuggingFace, Vertex AI Gemini, Ollama |
| Databases | Database connections | PostgreSQL, MySQL, SQL Server, MongoDB, Redis / Valkey, Elasticsearch |
| Voice | Speech services for voice agents | Deepgram, OpenAI Whisper, Azure Speech, ElevenLabs, OpenAI Realtime, Google Gemini Live |
| Image & Video | Image and video generation models | Provided by qRaptor — see below |
| Integrations | Everything else | Salesforce, HubSpot, ServiceNow, Jira, Slack, Microsoft Teams, Twilio, SendGrid, GitHub, Azure DevOps, AWS S3, Azure Blob, Gmail, Google Calendar, SMTP, REST API, webhooks |
Image & Video is provided by qRaptor. Image and video generation models come with your plan rather than being something you connect yourself, so that tab lists platform models only.
Two sources, one list
Every configuration in the list is either yours or provided by qRaptor, and both appear together under whichever category they belong to — a platform Claude model is a model, so it sits under Models alongside your own.
Use the source filter to separate them:
- All sources — everything.
- This project — only the configurations your team created. These are editable.
- Provided by qRaptor — the platform’s own models, included with your plan. These are read-only.
See Models provided by qRaptor for what comes with your plan.
Referencing a configuration
Anywhere a value can be templated — a tool’s HTTP header, an auth field, a Python handler, a node setting — you reference a configuration instead of typing a value:
{{config.<type>.<name>.<field>}}Four segments, in order:
| Segment | What it is | Example |
|---|---|---|
config | The namespace. Always this. | config |
<type> | The template the configuration was created from | postgresql |
<name> | The name you gave this configuration | orders_db |
<field> | The field you want from it | password |
So a Postgres connection named orders_db exposes {{config.postgresql.orders_db.host}}, {{config.postgresql.orders_db.password}}, and so on. The configuration’s name is shown on its card, and the create form displays the exact prefix to use while you’re naming it.
Leaving the field off — {{config.postgresql.orders_db}} — resolves to the whole configuration rather than one value, which is occasionally what you want when passing a connection somewhere wholesale.
In the tool builder you don’t have to type this by hand — type {{ to open the variable picker and choose from the Configs tab, which inserts the full reference for you. See Configuration & secrets for the tool author’s side of this.
{{system.…}} is a different namespace — it holds run context like system.runId and system.user.email, not configurations. Platform-provided models aren’t referenced by template at all; you pick them from a model selector or set them per feature in LLM Settings.
What’s in this section
- Add a configuration — pick a category, pick a template, fill in the fields.
- Secrets & encryption — how credentials are stored, and what happens when you edit one.
- Manage configurations — search, filter, edit, disable and re-enable.
- Models provided by qRaptor — what comes with your plan.
Next
- Add a configuration — create your first one.