SchemAxis is a web-based dependency-analysis, impact-analysis and data-lineage platform for Oracle databases. It reads schema structure through an open-source, read-only connector that you install and control inside your own network. It never reads the rows inside your tables, and it never receives your database credentials. The analysis itself is computed by a deterministic engine, not generated by a model. Everything below states how that is enforced rather than promised.
Read the connector's source (opens in a new tab)
This page is the same source the product's in-app Help & Security section renders.
The database username and password are configured locally in the local connector agent, on your machine or server. They are never sent to or stored by SchemAxis's servers (sensitive fields in the connector's local profiles are encrypted with AES-256-GCM using a key generated on your machine).
SchemAxis never writes to your database. The local connector agent inspects every statement before executing it and refuses anything that is not a read (only queries pass, while statements that modify data, lock rows or embed PL/SQL are refused, verified inside your network, regardless of what the cloud side asks). For defense in depth we also recommend running the connector with a read-only database account, then your database itself enforces it too.
The product reads metadata: table layouts, code, dependencies, index definitions. It never reads the rows inside your tables. Schema structure and code are fetched on demand for each analysis, processed in memory, and discarded. The platform does not keep a copy of your schema or source code. What persists is only your own workspace: analysis history, chat conversations, comparison fingerprints, and documents you upload.
The local connector agent dials out to the platform over an encrypted tunnel (WebSocket over TLS, authenticated with a short-lived signed token you can revoke at any time). You never open an inbound port, and closing the connector immediately severs all access.
Secrets stored on the platform on your behalf, such as your AI provider key, are encrypted at the application layer before being written to the database (Fernet/AES on top of disk encryption). Passwords are never stored, only slow one-way hashes (bcrypt).
Every record carries your organization's identity, resolved from the verified login token on every request and enforced at the database layer via RLS. One customer can never list, read, or touch another customer's connections, chats, documents, or results.
AI features are BYOK (bring your own key): prompts go to the provider you configured with the key you own, billed to your account. SchemAxis sends the AI only schema metadata and code structure (never table rows and never your database credentials) and trains no model on any of it. Because the provider account is yours, the retention and training terms that apply downstream are ones you hold and can verify directly, including a model running entirely inside your own network.
The dependency graph, the impact list, the lineage trace and the environment diff are produced by a deterministic engine that parses your data dictionary and your PL/SQL source. The same schema and the same question produce the same answer, and every object named in a result exists in your catalog. AI is an optional layer on top of a finished deterministic result: it explains, escalates the cases the engine itself flagged as ambiguous, writes documentation and drafts code. Its claims are validated against real dictionary objects before being displayed. With no AI key configured, every deterministic feature works normally.
Security-relevant actions are recorded in an audit trail visible to your admins. If you delete your organization, the data is physically removed (hard delete), not just hidden.
SchemAxis's application servers and PostgreSQL database run in the European Union. The full processor list, the legal bases, and the retention periods are in the Privacy Policy: that document is the authoritative one.
SchemAxis holds no SOC 2, ISO 27001, or HIPAA certification, and does not claim one. The product is in closed beta: the controls described on this page are implemented and testable today, formal third-party attestation is not part of this phase. Ask us for the architecture detail your review needs and we will answer it directly.
Security reports are welcome. Write to the contact address published on this site with the affected endpoint or component and how to reproduce the finding; please do not test against other customers' data or run denial-of-service traffic against the platform.
The Terms of Service and the Privacy Policy are versioned documents, published on the application domain and referenced by the acceptance record on every account. They are linked here rather than reproduced.
SchemAxis is a web-based dependency-analysis, impact-analysis and data-lineage platform for Oracle databases. It reads schema structure (tables, views, packages, triggers and their dependencies) through an open-source read-only connector you install inside your own network, and answers questions such as "if I change this column, what breaks?" It is built for the developers, DBAs and system integrators who maintain legacy Oracle databases.
Oracle only. SchemAxis reads the Oracle data dictionary and parses PL/SQL, and that depth is what answers the questions the tool exists for. The direction is deeper into the Oracle estate (the applications and integration jobs built on top of the database), not sideways: no other database engine is on the roadmap now. It works with Oracle Database 11g Release 2 (11.2) and every later release, including 19c and 23ai. The connector reaches your database the way a SQL client on that machine would, by service name or SID. The install guide shipped with it lists the releases tested and what it can connect to today.
Yes, and it is one of the main reasons the product exists. Large Oracle applications assemble statements at runtime (EXECUTE IMMEDIATE, dynamic cursors, table names composed from strings) and the data dictionary records none of those references, so a dependency list built from the catalog alone is silently incomplete. SchemAxis parses the PL/SQL source itself to recover the object names those statements resolve to, and reports the ones it could not resolve as an explicit gap rather than omitting them: a statement whose target could not be confirmed is shown as a warning on the result, not hidden. The ambiguous cases can then be escalated to an AI Deep Scan, which reads the surrounding code and reconciles the runtime targets against objects that really exist in your catalog.
The analysis is computed, not generated. Dependencies, impact, lineage and environment differences come from a deterministic engine that reads your data dictionary and parses your PL/SQL: the same schema and the same question produce the same answer every time, and every object in a result is one that exists in your catalog. AI is a separate, optional layer that runs on top of a finished deterministic result. The AI Deep Scan explains the result, resolves the cases the engine itself flagged as ambiguous (such as dynamically built SQL); AI assistant writes documentation and drafts code. Every claim it makes is checked against real dictionary objects. You can turn the AI off.
No. The local connector agent only makes outbound connections, like a browser. If your network allows outbound HTTPS, the connector works.
No. The local connector agent blocks every non-read statement before it reaches your database, inside your own network. Running it with a read-only database account adds a second, database-enforced guarantee.
There is an important difference between the two. To run an analysis, schema structure (object names, column definitions, code, dependency information) is fetched on demand through the local connector agent's tunnel, processed in memory, and discarded: we keep no copy of your schema or your code. Never table contents, never your database credentials. What we do store is your own workspace: your analysis history (which tool ran, on which object), your AI conversations, environment-comparison snapshots (stored as content fingerprints, not raw code), and any documents you explicitly upload. All of it is yours, scoped to your organization, and physically erased if you delete it.
No. AI features need an API key from a provider of your choice (BYOK), and several reputable providers hand out a key in minutes with a free tier that is plenty for evaluation. Without a key every non-AI feature (dependency graphs, impact analysis, lineage, schema comparison) works normally.
AI is bring-your-own-key: SchemAxis acts as a secure relay to the provider you configured, under your key and your billing. The AI receives only the schema metadata and code context needed to answer, never business data. We don't retain those payloads beyond the conversation you see in your chat history, nor use them for anything else, nor train any model on them. The provider account is yours, so its own retention and training terms are ones you accept and can verify directly; if that is still too much exposure, the same features run against a model hosted inside your own network. Without a configured key, AI features are simply off.
Stop the local connector agent's process on your machine, since it is the only path to your database, all access ends immediately. Alternatively, revoke or regenerate its token from the Connections page: a revoked token disconnects a live connector within seconds.
Nothing. Design partners use the full product free of charge, with no credit card and no obligation, and pricing is announced before general availability.
Access is by invitation during the closed beta. Submit your email address through the request form on this site; eligible requests receive an access code by email, which is what unlocks account creation. The only data the form requires is the email address the code is sent to.