Integrations — Model providers
NerveStax and Anthropic
Bring your own Anthropic key. NerveStax has no fallback key.
What we do
What NerveStax does with Anthropic
Add your organization's Anthropic API key and NerveStax runs its agents on it. There is no shared key behind the product: with no provider configured, agents refuse to run. Keys are stored encrypted, referenced by handle, and never shown back, and calls go to Anthropic under your agreement with them, not through us.
-
Your key, per organization
Each organization adds its own key. NerveStax ships no fallback key, so if none is configured the agent stops with a clear error rather than quietly billing someone else.
-
Choose the model per agent
Different agents get different models: a small one for routine context lookups, a larger one where a change is being designed. You set the model, not the product.
-
Calls are never proxied
Requests go from your NerveStax deployment to Anthropic. We do not sit in the middle of them, and the product ships no third-party analytics or error-tracking SDKs.
-
Spend tracked per call
Every model call is recorded with the organization, workspace, user, conversation, agent, model and token counts, so you can see where the bill comes from before the invoice does.
-
Prompt caching where supported
Long-lived context such as your dbt manifest and repository files is cached across a conversation's turns where the provider supports it, which is most of what keeps cost down.
Background
About Anthropic
Anthropic builds the Claude family of models and sells access to them through its API. Teams buy capacity directly and hold their own key and their own agreement.
How it connects
What connecting it looks like
-
01
Create a key in Anthropic
Make an API key in your Anthropic account, ideally one scoped to this use so it can be rotated on its own.
-
02
Add it under AI providers
Paste the key into Settings. It is encrypted at rest with a key held outside the database, and no screen shows it again.
-
03
Pick a model per agent
Assign models to agents. Start small for context and triage work, and raise it where you find the reasoning is not enough.
-
04
Watch the token usage page
Check spend by agent and workspace after the first week, and adjust the model assignments from what you see.
Scope
What this does not cover
- What reaches Anthropic: schemas and table metadata, dbt artifacts, repository files, command output and Airflow logs with secrets redacted, alert payloads, and, when an agent queries or samples a table, the rows it retrieves.
- Secrets are redacted from what the model reads. Result rows are not masked, so treat your model provider as in scope for whatever an agent is allowed to query.
- Rotating or revoking the key is done in NerveStax by replacing it; there is no endpoint that reveals the stored value.
Questions
Anthropic and NerveStax
Q01Does NerveStax see my model key?+
The key is stored encrypted with AES-256-GCM under a key held outside the database, referenced by a handle, and no endpoint reveals it; screens show a masked value. Self-hosted, both the database and that encryption key are entirely yours. Calls go from your deployment to Anthropic, never proxied through us.
Q02What exactly is sent to Anthropic?+
Schemas and table metadata, dbt manifest, lineage and run results, repository file contents, command output and Airflow logs with secrets redacted, uploaded documents, alert payloads, and the rows an agent retrieves when it queries or samples a table. It goes to Anthropic under your key and your agreement with them.
Q03Do you use a key of your own if ours is missing?+
No. There is no platform fallback key. With no provider configured, the agent raises a configuration error and does nothing. That is deliberate: it means no work can ever run on someone else's account or outside your agreement with a provider.
Q04Can we set spending limits?+
Limits live in your Anthropic account, where they belong, since the key is yours. What NerveStax gives you is attribution: an append-only record of every call with its tokens, model, agent, workspace and user, so you can see which agents and which teams drive the cost.
Q05Can we switch providers later?+
Yes. The provider is configuration, not a rebuild. Anthropic, OpenAI, Azure OpenAI and any OpenAI-compatible endpoint are all supported, and you can move by adding the new provider and reassigning models to agents.