Q&A with full platform knowledge
Full RAG access to Etlworks documentation and the website. Ask anything — “how do I configure CDC for PostgreSQL?” — and get a real answer, not a keyword-search result dressed up as AI.
Simba · the Etlworks AI agent
Build data flows by describing what you need.
Simba is the agent built into Etlworks. Tell it what you’re trying to do — “load Stripe charges into Snowflake daily,” “sync Salesforce contacts to HubSpot” — and it builds the flow. You stay in control; Simba does the typing.
- Build pipelines from a plain-English description
- Debug failing flows — reads errors, schemas, configs, and run logs
- Optimize slow pipelines — parallelization, batching, pushdown
- Migrate logic from SnapLogic, Talend, Airflow, or custom Python
- Operate your environment — flow status, lag, recent errors, metadata search and updates
- 200+ flow types Simba can build
- 3,979 templates Simba reaches for
- 270+ connectors
- BYOK use your own OpenAI key
The hard part
You shouldn’t have to know the platform to use it.
Most data integration tools require knowing the platform’s vocabulary before you can build anything. Which flow type? What template? Which connector? How does mapping work? That’s a learning curve before you ship anything.
What Simba does
Seven things Simba can do today. Most of them are about turning English into working data pipelines — the rest are about answering questions about the platform and the data inside it.
Write code where you write code
JavaScript, Python, SQL (multiple dialects), and the Etlworks CLI. Simba knows whether you’re in a code editor, SQL editor, or CLI console — and writes in the right language and dialect. Hundreds of built-in transformation functions on tap.
Build flows live in the canvas
“Load Stripe charges into Snowflake daily, dedupe on charge_id.” In Composer, the canvas updates as you describe what you want. Source, destination, mapping, schedule — assembled in front of you, ready for approval before anything runs.
Build connections to APIs it’s never seen
For HTTP APIs, Simba searches the web for the API’s documentation and configures the connection automatically: endpoints, methods, headers, auth, pagination. “Create a connection to the PayPal endpoint that returns stored payment methods” — and it does.
DevOps and DataOps work
System health, flow stats, log inspection, error diagnosis, metadata search, metadata updates. Show Simba a failing flow; it reads the run log, finds the cause (schema drift, auth expiry), proposes a fix, and applies it on approval.
Mapping and Explorer awareness
Content-aware and location-aware. Knows the columns in the mapping, in the source, and in the destination. Knows the SQL dialect and the language you’re working in. Helps you where you are, not where it wishes you were.
Data work in Explorer
Sees tables, views, endpoints, columns, datasets. Samples data and runs SQL on its own to answer your questions — no copy-pasting between tabs, no context-switching to a separate query tool.
Three ways to use Simba
Same agent, three surfaces. Pick whichever fits the moment.
In the product
Inside Composer
Chat with Simba while building flows on the canvas — and watch the canvas update live as you describe what you want. Drop two connections and ask Simba to wire them, or describe the whole pipeline and have it built in front of you. The default way most teams use Simba.
See ComposerConversational
Refine in plain English
“Add a daily summary email.” “Filter out test rows before loading.” “Run this only on weekdays.” Simba updates the flow and shows you what changed before applying.
Watch a demoProgrammatic
Agent API
Drive Simba from your own AI stack. REST API, Python and Bash clients. Use as a subagent in LangChain, CrewAI, or AutoGen. Streaming, multi-turn sessions, direct tool access.
Read the API docsWhat Simba is and isn’t
Honest about both. Simba is good at most things and a few things it deliberately leaves to you.
Simba is good at
- Building flows for known source-destination patterns (Salesforce → Snowflake, etc.)
- Building connections to HTTP APIs by reading their docs from the web — endpoints, auth, pagination, headers, all of it
- Translating SQL, Python, JS, and most pipeline definitions into Composer flows
- Reading schemas and inferring field mappings automatically
- Diagnosing flow errors from run logs and suggesting fixes
- Writing inline transformations in JavaScript, Python, or SQL — across multiple dialects, with hundreds of built-in functions
- Searching and updating metadata across your environment
Simba isn’t trying to
- Translate proprietary closed-source formats — Informatica’s PowerCenter mappings, some legacy ETL XMLs, vendor-specific binary formats. Simba can take a description and rebuild the logic, but won’t parse formats we don’t have access to.
- Replace your data engineer for complex orchestration — multi-system distributed transactions, regulatory-grade audit chains, novel architectures with no precedent. Simba helps; it doesn’t lead.
How Simba handles your data
Simba runs on the same Etlworks infrastructure your flows do. Same data isolation, same SOC 2 controls, same retention policies.
No training on your data
We use frontier LLMs from major providers under enterprise agreements that prohibit training on customer data. When you bring your own OpenAI key, your data flows through your provider relationship under your own terms — Etlworks doesn’t sit in the middle. When using the Etlworks-managed wallet, your conversations aren’t shared across customers and aren’t used for training.
On/off and access controls
The agent is enabled per organization — all-or-nothing, no partial agents. Inside an organization, role-based access controls determine which users can do which things; the agent operates as the calling user, never above their permissions. (Some tools, like web search, have additional internal restrictions managed by Etlworks.)
Approval before changes
Simba shows you what it’s about to do before applying. Every flow change, every config edit, every credential update — you approve. Optional auto-apply for low-risk operations.
Full audit trail
Every Simba action is logged: who asked, what it built, what changed, when. Available alongside human-driven changes. Connection credentials and secrets never leave the runtime — the agent references them by name but never receives raw values.
See security detailsFAQ
Do I have to use Simba?
No. Composer supports drag-and-drop, code, and CLI building without Simba ever entering the picture. Simba is opt-in. Most teams use it for the parts they don’t already know how to build, and skip it for the parts they do.
What model powers Simba?
Etlworks runs Simba on the latest frontier OpenAI models — we pick and update them as new ones ship. Whether you bring your own key (BYOK) or use the Etlworks-managed wallet, you don’t choose the model; you always get the current best.
The intelligence sits in the toolchain wrapped around the model: knowledge base search, template lookup, schema introspection, CLI execution, and flow validation. The model handles language; the toolchain handles the platform-specific work.
Can I use Simba in my own AI app?
Yes. The Agent API exposes Simba as a REST endpoint with Python, Bash, and PowerShell clients. Use it as a subagent in LangChain, CrewAI, AutoGen, or any orchestration framework. Full developer docs
Does Simba work in on-premise deployments?
Yes. Simba ships with every Etlworks deployment — cloud, hybrid, and on-premise. For air-gapped on-prem environments, the language model can be configured to use a self-hosted inference endpoint instead of a hosted provider.
How is this different from generic AI assistants like ChatGPT or Claude?
Generic assistants don’t have access to your Etlworks platform. Simba does. It can read your existing flows, search 3,979 templates, inspect schemas in your connected systems, run CLI commands, and apply changes — all inside the Etlworks platform with your permissions and audit trail intact. A generic assistant can write you SQL; Simba can ship you a working pipeline.
How is this different from Matillion Maia or Informatica CLAIRE?
Maia and CLAIRE are real efforts at agentic ETL. Differences: (1) Etlworks’s agent is exposed as a subagent for your own AI stack via REST API — Maia and CLAIRE are platform-internal. (2) Simba can build connections to HTTP APIs it’s never seen by reading their docs from the web — endpoints, auth, pagination, headers. (3) Etlworks ships AI features outside the agent too — auto-mapping, in-app search, prompt-driven insights — covered on our AI overview page. (4) BYOK is supported: bring your own OpenAI key and pay nothing to Etlworks for inference.
What does Simba cost?
Most teams bring their own OpenAI API key — Simba uses it directly, and you pay your AI provider, not Etlworks. Configure once per organization, or per user if you want individual billing. There’s no Etlworks markup or middleware fee on the AI usage.
If you’d rather not manage your own provider relationship, Etlworks offers an AI wallet with auto-recharge — fund it once, set a monthly cap and a low-balance recharge trigger (the same model OpenAI uses directly). Wallet usage is billed at-cost.
Every plan also includes a small monthly Simba allowance for trying things out without configuring either option. See pricing
Try Simba.
14-day free trial includes Simba credits. Build your first flow by describing what you need. No credit card.