Head-to-head · updated 13 September 2026

DataShield vs Castor: the catalog documents the data, but who governs the agent reading it?

First, the name. Castor became CastorDoc, and on 19 March 2025 Coalesce bought it and renamed it Coalesce Catalog. If you searched for Castor and landed on a transformation vendor's site, that is why. The product is good at what it always did well. Point it at a warehouse and it writes descriptions for every table and column, maps column-level lineage, and lets a business user ask a plain English question instead of filing a ticket. Gartner made Coalesce a Visionary for metadata management in 2025. The hero line today is "Simplify data discovery for all data people," and they earn it.

We are not a discovery tool for your whole estate, and we will not pretend otherwise. DataShield Ontology is a governed data plane with a catalog inside it. It scans and profiles a live PostgreSQL source in place, labels columns against 129 PII, PHI, financial and secret classes, and stores no column values by default. Datasets are tokenized at ingest; agents query tokenized data over MCP; detokenization is a privileged, audited operation. It runs on your own hardware. Below is the honest split, including the rows Castor wins.

Get free Demo Center access Get an instant quote

The short version

Pick DataShield when

The catalog has to run inside your own network. Coalesce Catalog is SaaS. Private networking shows up only on the top tier, and no self-hosted edition is documented.
You need the tool to find the PII nobody tagged. We label columns against 129 classes, including all 18 HIPAA Safe Harbor identifiers, using regex, checksums and column-name evidence. No model, so the verdict is reproducible.
An examiner will ask you to prove an agent's access log was not edited. Our chain answers with math. Run the verifier.
You want the data itself protected, not just labelled: tokenized at ingest, generalized on the way out, erasable by destroying a key.

Pick Castor when

You want every asset documented by Monday. Their AI description coverage is the best part of the product and we do not compete with it.
You need column-level lineage across Snowflake, BigQuery, Databricks, Fabric and your BI tools, stitched end to end. Ours is typed and access-gated, but narrower.
You already run Coalesce Transform. Catalog intent flows into the pipelines and quality results flow back. That loop is real and we have nothing like it.
The buyer is a business team, not a security team. Their search and assistant are built for people who do not write SQL.

Bottom line: Castor tells your people what the data means. DataShield decides what an agent may do with it and keeps proof of the decision. Most teams who buy us keep a catalog like theirs for the rest of the estate.

Feature by feature: AI data catalog against governed data plane

Competitor cells describe what Coalesce's public site, docs and changelog say as of the date above. If we have mischaracterised something, email support@myorg.ai and we will correct it, credited.

What matters	DataShield	Castor	Edge
Catalog coverage and connectors	We register a provider over a live connection, then scan and profile it in place with no rows leaving the source. That is PostgreSQL today. Snowflake, BigQuery, Databricks, S3, Salesforce and Kafka are declared on the roadmap with no handler yet, and we say so in the config file.	Snowflake, BigQuery, Databricks, Microsoft Fabric, Redshift in private preview, plus PostgreSQL, SQL Server, Tableau, Power BI, Looker, dbt and Fivetran. Much wider than us.	◇
AI documentation	We do not auto-write prose descriptions. Our glossary is materialized from entity types as ISO/IEC 11179 data element concepts, which is precise and dull.	The flagship feature. AI-generated table, column and dashboard descriptions, with admin-set company context feeding the generator since October 2025.	◇
Lineage	Typed lineage over a fixed predicate vocabulary, traversable with per-hop access gating, so a user only walks edges they may see on both ends. It is derived from the pipelines that own the data, not a stored column graph.	Column-level and asset-level lineage, Snowflake Dynamic Tables support, and cross-warehouse indirect lineage stitching added in May 2026. Deeper than ours.	◇
Sensitive data classification	129 built-in field classes covering PII, PHI, financial data and secrets. Regex plus checksum validation (Luhn, NPI, Verhoeff, ABA, IBAN, GTIN), column-name lexicons and anti-pattern suppressors. Per-class severity, confidence and regulation reference. Deterministic, so two runs give the same verdict.	Manual tagging. Bulk column tagging landed in July 2025 to make it faster. We found no automated PII or PHI classifier in their public material.	◆
Sample values in the catalog	Default policy is omit: no column value is copied into the catalogue at all. Masked stores shape only. Plain requires an explicit acknowledgement and writes an audit row, and an operator can see exactly how many columns hold cleartext.	Not described in their public docs. Worth asking, since a catalog that stores previews of a PHI column has quietly become a second copy of it.	◆
Tokenization and masking	Datasets are tokenized at ingest; agents query tokenized data over MCP; detokenization is a privileged, audited operation. On top of that, quasi-identifier generalization: dates to year, decade or age band, ZIPs to 3 or 4 digits, partial phones, SSNs and emails, each with a measured cardinality-reduction score. These are features you switch on, not defaults.	Tagging and access controls. No masking, tokenization or anonymization of values found.	◆
MCP and agents	More than 200 MCP tools across Ontology, Auth, Corpus and Lighthouse. Agents do not just read metadata, they query the governed data and act on it under tool tokens with scope ceilings.	A Catalog MCP server since December 2025 with 8 read tools, lineage tools added May 2026, and OAuth with per-user RBAC since September 2026. Static API tokens for MCP are now deprecated. Good hygiene, read-only scope.	◆
Agent authorization	Every governed tool call passes a scope ceiling, an authority tier and a live revocation re-check before dispatch. Revoke an agent and the next call fails closed, not the next token refresh.	OAuth plus the platform's role-based access control for the signed-in user. Copilot has RBAC and audit logging. We found no per-call authority re-check.	◆
Audit evidence	SHA-256 hash chain with Ed25519-signed checkpoints that are themselves chained. Verification names the failure: tampering, insertion, deletion or truncation. Try the verifier.	Audit trails for the data lifecycle, and audit logging in Copilot. We found no published tamper-evidence mechanism.	◆
Break-glass and erasure	Scoped, time-boxed emergency access for agents that auto-revokes. GDPR erasure by crypto-shred of per-subject key material, plus ISO 27560 consent receipts. The chain still verifies afterwards.	Not described in their public material.	◆
Deployment	Self-hosted in your own cloud or data center, or a dedicated single-tenant server we operate. Docker images ship for Auth, Ontology, Corpus and Lighthouse, with a signed deploy manifest Guardian verifies. Ed25519 audit-signing keys can live in your KMS or HSM. HMAC tokenization keys sit in your environment today, not in a KMS.	Multi-tenant SaaS. Private Networks appear on the Business Critical tier, along with a BAA. No self-hosted or on-premise edition is documented.	◆
Maturity and pricing	Live in production, Guardian and Lighthouse since April 2026. SOC 2 not yet certified, and we say so. Published pricing and a scoped instant quote.	Backed by a funded parent, Gartner Visionary in 2025, and two acquisitions absorbed. Developer tier free, Starter listed at $150 per user per month billed annually, with Catalog gated by integration count rather than sold on its own.	◇

◆ DataShield leads◇ Castor leads◈ comparable

Castor claims are drawn from coalesce.io, castordoc.com and the Coalesce Catalog changelog, last checked 13 September 2026. We link them below rather than work from memory.

Three things you get here that you won't get from an AI data catalog

Labels the machine works out for itself

A tag is only as good as the person who remembered to apply it. We check values and column names against 129 classes, validate the ones that carry a check digit, and suppress the lookalikes. The result is reproducible from a config digest, because there is no model in the path to drift. How the catalog works.

Authority that can change mid-flight

An analyst leaves on a Friday. Their agent is 20 minutes into a 40-minute job. With DataShield the next governed tool call is re-checked against current authority and fails closed. A read-only assistant would finish the job and tell you about it later. How Auth does it.

Proof that survives an audit

A log that can be edited proves nothing. Ours is a hash chain with signed checkpoints, and the verifier says what broke, not just that something did. That is the property EU AI Act Article 12 and HIPAA §164.312(b) reviewers care about. Try it in your browser, no signup.

Where Castor is genuinely stronger

We would rather you heard this from us. The CastorDoc team solved the boring, brutal problem of getting a catalog documented at all. Their AI writes descriptions across the estate, their search answers a plain question, and their lineage now stitches across warehouses and into BI tools. Since the Coalesce acquisition the story got tighter, not weaker: business intent defined in the catalog is enforced in the transformation pipelines, and quality results from Coalesce Quality land back in the catalog. If your estate is Snowflake-shaped and your problem is that nobody knows what dim_cust_v3 means, buy theirs. Their MCP work is sensible too. OAuth with per-user RBAC, shipped September 2026, is ahead of most catalog peers, and deprecating static tokens was the right call.

Here is the push-back. Their MCP server hands an assistant metadata under the signed-in user's permissions. That is a good read design, and it is the whole design. There is no per-call authority re-check, no break-glass, no tamper-evident record of what the assistant did, and nothing that touches the values themselves. Catalog governance is tagging plus access control, which assumes someone already knew the column held a national ID. And all of it lives in their tenancy. If your rule is that the metadata about patient records cannot leave your building, the conversation ends before the demo starts. That is the seam we sell into, and it sits underneath a catalog like theirs quite happily.

Questions worth asking both of us

These are the questions we would want answered if we were the ones buying. Ask them on every call, ours included.

Can you cryptographically prove an audit log entry wasn't deleted?

DataShield: yes. Each record commits to the one before it, checkpoints are signed and chained, and verification tells deletion apart from truncation and from tampering. Try it at /verify. Castor: their site describes audit trails for the data lifecycle, and Copilot logs its actions. We found no tamper-evidence mechanism in their public docs. Ask them to show one.

What happens to a revoked agent mid-session?

DataShield re-checks authority on every governed tool call, so revocation lands on the very next call. Castor's MCP server signs the user in over OAuth and applies the platform's role-based access control per request. That is better than a static token. Ask how fast a permission change reaches an assistant that is already mid-task, and what record you get of the calls it made before the change.

How does GDPR erasure interact with the audit trail?

DataShield crypto-shreds the per-subject key material and issues an ISO 27560 consent receipt. Actor identities in the chain are HMAC-committed, so the evidence still verifies once the subject is gone. Castor documents no erasure mechanism we could find. Worth asking, because a catalog that stores sample values or AI-written descriptions of a person's record now holds copies of it too.

I searched for a Castor alternative. Is that even the right search now?

Probably not, and here is the honest answer. Castor became CastorDoc, Coalesce bought it in March 2025, and it is now Coalesce Catalog, sold with Coalesce Transform and Coalesce Quality. The old castordoc.com still loads with a banner saying the name changed. So the real question is whether you want a catalog bundled with a transformation tool. If yes, theirs is a strong bundle. If you wanted the standalone catalog you first shortlisted, look at us, Secoda or Select Star and compare on what you actually need.

Can Coalesce Catalog run in our own data center?

Not that we can find. Their pricing page lists Private Networks and a BAA on the Business Critical tier only, and no self-hosted or on-premise edition appears in their docs. DataShield ships Docker images for Auth, Ontology, Corpus and Lighthouse, with a signed deploy manifest that Guardian verifies, and runs self-hosted in your own cloud or data center or as a dedicated single-tenant server we operate. If self-hosting is a hard rule for you, ask them in writing early.

Does DataShield have SOC 2?

Not yet, and we will not imply otherwise. Auth is live with a public threat model and a verifier anyone can run. Guardian and Lighthouse have been in production since April 2026. Design-partner terms include source escrow, so a small vendor is not a single point of failure. Details on the security page.

Coalesce Catalog's current hero: "Simplify data discovery for all data people. Discover, centralize, and govern your data assets with help from AI." — coalesce.io, 13 Sep 2026
Coalesce acquired CastorDoc and relaunched it as Coalesce Catalog, "an intuitive, AI-driven metadata management solution for modern data teams." — Coalesce company news, 19 Mar 2025
Catalog MCP server launched with 8 read tools (8 Dec 2025); lineage tools added (15 May 2026); OAuth with per-user RBAC added and static API token auth for MCP deprecated (7 Sep 2026). — Coalesce Catalog changelog, 13 Sep 2026
Pricing: Developer free, Starter $150 per user per month billed annually with 3 Catalog integrations, Enterprise custom with 5, Business Critical adds Private Networks and a BAA. No self-hosted edition listed. — coalesce.io/pricing, 13 Sep 2026
Coalesce acquired SYNQ and launched Coalesce Quality, completing the Transform, Catalog and Quality "data operating layer." — Coalesce company news, 9 Mar 2026
Coalesce named a Visionary in the 2025 Gartner Magic Quadrant for Metadata Management Solutions. — Coalesce company news, 24 Nov 2025

Other head-to-heads

Catalog

Get free Demo Center access

You've seen the proof

Ready for a number? Scope your deployment and we'll price it against your own economics.

Get your quote →