Agentic AI for data teams — dbt · airflow · the warehouse
Ship dbt changes you didn’t write.
NerveStax keeps a live map of every model, column and run in your platform — then does the work in a real sandbox and opens the pull request.
Below: one real run, start to finish. Nothing merges without your approval.
>
- read manifest.json — 412 models, 1,208 columns
- trace lineage — mart.orders_daily ← 14 downstream assets
- source landed 6h late on 2 of the last 7 runs
- edit models/staging/src_orders.yml
- run dbt build --select stg_orders+
- Runs in
- An isolated sandbox, one per conversation
- Ships via
- A pull request on your own branch
- Remembers
- Assets, jobs and runs — indexed, not re-parsed
- Secrets
- Encrypted references, never in a prompt
The gap
A coding agent sees files.
Your platform is not a folder.
Three things a repo-only agent structurally cannot know — and all three are why you end up doing the work yourself.
It re-reads your repo every session
Parse the project, answer one question, throw it all away. Compile-time truth, re-paid hourly, with no memory of what you asked yesterday or what it changed last week.
It can’t see what actually ran
A file says what should exist. It says nothing about last night’s failure, the model that got 4× slower, or the table that quietly drifted from its declared schema.
It can’t prove the change works
Plausible SQL is not tested SQL. Without a warehouse and a real dbt build,
the person who finds out is you, in production, at 7am.
It runs the code
before you do.
Every conversation gets an isolated sandbox with your repo checked out. The agent edits
files, runs dbt build and dbt test against a real target, reads the
failures, fixes them, and runs again — then pushes a branch and opens a PR.
Your git history stays the source of truth. The sandbox is a cache, and it’s disposable.
$ dbt build --select stg_orders+ --target dev 14:02:04 Found 412 models, 1,208 tests, 39 sources 14:02:07 1 of 15 START sql view model stg_orders ........ [RUN] 14:02:08 1 of 15 OK created sql view model stg_orders ... [OK 0.94s] 14:02:09 2 of 15 OK source_freshness src_orders ........ [PASS] 14:02:13 15 of 15 OK created table model mart.orders_daily [OK 3.1s] 14:02:13 Done. PASS=15 WARN=0 ERROR=0 SKIP=0 TOTAL=15 $ git push -u origin nx/orders-freshness branch pushed · PR #284 opened
It remembers your
whole platform.
One graph fusing what should exist with what actually happened: models, columns, tests and lineage from your repo, joined to run status, duration, row counts, freshness and schema drift from the warehouse.
Indexed once and kept current — so every question starts from what your platform is really doing, not from a cold re-parse.
It stops and
asks you first.
Writes, merges and anything destructive pause at an approval gate and wait for a human. The run doesn’t restart when you answer — it resumes exactly where it stopped, with its working state intact.
Credentials are never inlined into prompts, code or logs. They stay encrypted references that resolve at execution time, inside the sandbox.
tool open_pr(branch="nx/orders-freshness", base="main") risk write · requires approval diff 2 files changed, +18 −0 creds secret_ref://github/analytics-pat (resolved in-pod)
It shows you
the bill.
Token spend broken out by workspace, agent and conversation, so AI cost is a line you manage instead of a monthly surprise. Bring your own provider keys and choose which model each agent runs on.
Getting started
Connect once. Ask in English.
Review a diff.
Connect your stack
Point NerveStax at your dbt repo, your warehouse target and your Airflow cluster. Credentials go into the secret store as references — never into a prompt.
Ask for the outcome
“Model this new source.” “Why did last night’s run fail?” “Add tests to everything
downstream of stg_orders.” It already knows your models and your run history.
Review the pull request
A branch, a diff, passing test output and a written rationale. Approve it or send it back with a comment. Nothing reaches production without your merge.
Security
Built for a team that
answers to an audit.
- Execution
- Each conversation runs in its own isolated sandbox with scoped credentials. One agent’s run cannot reach another’s.
- Credentials
- Warehouse and git secrets live encrypted in a secret store and resolve at execution time — never inline in prompts, code or logs.
- Source of truth
- Work lands as branches and pull requests in your own git remote. Sandboxes are caches and get thrown away.
- Production
- Agents work against a development target. Merging is yours, behind your existing CI and review rules.
- Model access
- Bring your own provider keys and pick the model per agent. Spend is itemised per workspace and conversation.
- Deployment
- Hosted by us, or self-hosted in your own cloud account — the whole stack ships as containers.
Questions
Everything people
ask us first.
Q01What exactly is NerveStax?−
An agentic workspace for data teams. Connect your dbt repo, warehouse and orchestrator; NerveStax builds and maintains a model of your platform, and its agents use that context to do real work — writing models and tests, debugging failed runs, tracing lineage — with every change arriving as a reviewable pull request.
Q02Which warehouses and tools do you support?+
dbt is first-class, with Snowflake, Trino, DuckDB/MotherDuck and Postgres targets working today, and Airflow alongside it. The adapter layer is deliberately pluggable — if you run something else, say so in the form below and we’ll tell you where it sits on the roadmap.
Q03Does it write to production?+
No. Agents work in an isolated sandbox against a development target, and anything that changes your codebase arrives as a pull request. Merging — and therefore production — stays entirely in your hands, behind your existing CI and review rules.
Q04Does my code or data leave my infrastructure?+
Your repository is cloned into an isolated sandbox to run dbt, and model prompts include metadata such as schemas, column names and run results. We don’t ship warehouse rows to the model provider as a matter of course. If your policy requires everything to stay inside your own network, the self-hosted deployment is exactly that.
Q05Do I have to restructure my dbt project?+
No. NerveStax reads your existing project — manifest, run results and catalog — as it already is. If it builds on your machine, it builds here.
Q06Can I use my own LLM API keys?+
Yes. Add your provider keys in settings and choose which model each agent uses. Token spend is itemised per workspace, agent and conversation so you can see exactly where it goes.
Q07Is there a self-hosted option?+
Yes — the whole stack runs in containers and can be deployed inside your own cloud account. Pick “Question” in the form below and mention self-hosting; we’ll send the deployment notes.
Q08What does it cost?+
Pricing lands with general availability. The private beta is free, and beta teams get founding-customer pricing at launch.
Q09What’s coming next?+
Deeper Airflow authoring, ingestion pipelines, and a richer catalog surface on top of the context layer. Beta customers materially shape that order — that’s most of the point of the beta.