Head-to-head · updated 13 September 2026

DataShield vs Metaplane: your tables are fresh, but can you prove what the agent read?

Metaplane is a solid tool. It's honest about its job, too. Point it at Snowflake. Wait a day. It tells you when a table goes stale. Or when a row count drops off a cliff. Or when someone changes a column type. On a Friday. Column-level lineage shows you which dashboard breaks. There's a free tier, no card needed. That's rare, and it's welcome. Datadog bought them in April 2025. The product still sells on its own. It goes by Metaplane by Datadog now.

We do none of that work. DataShield has no anomaly detection. No monitors on your tables, and no freshness SLAs either. There's no issue queue. We sit one layer up. The datasets your AI agents read. Each dataset gets a full profile. Plus a trust score. We classify fields against 129 PII and PHI classes. We tokenize datasets at ingest. Agents then query tokenized data over MCP. Turning a token back into real data: a privileged step. Always logged. We check each tool call against the agent's access. First. The verdict then seals into a hash chain. You can verify it yourself. Below is the honest split. Wins for them, too.

Get free Demo Center access Get an instant quote

The short version

Pick DataShield when

Agents already query your data. Someone will ask you to prove what they touched. Our audit chain answers with a hash. Not a screenshot. Run the verifier.
You need to cut an agent's access mid-job. The next tool call should fail, right then. An alert on Monday is not the same thing.
The data has PII or PHI in it. You want those fields classified and tokenized. That happens before a model sees a row.
Nothing may leave your network. Not even metadata. We self-host on your own hardware.

Pick Metaplane when

Your problem is silent breakage. Stale tables. Dropped rows. A schema change nobody announced. That's their whole job. We don't do it at all.
You want to try it this afternoon. Free forever at 10 tables, no card. They claim setup in about 15 minutes. Anomalies flag within a day.
You live in Snowflake. There's a Native App. You can pay out of credits you've already committed.
You already run Datadog. Metaplane is Datadog's data observability line now. One vendor, one invoice, one on-call rota is a real argument.

Bottom line: Metaplane tells you the data broke. DataShield tells you who read it, under what access. It proves the record wasn't edited. These are different jobs. Most teams running agents in production will want both.

Feature by feature

Competitor cells describe what Metaplane's site says. And Datadog's docs. We use the date above. Did we get something wrong? Email support@myorg.ai. We'll fix it and credit you.

What matters	DataShield	Metaplane	Edge
Anomaly detection and alerting	We have none. We won't fake it. No monitors on tables, no freshness SLAs. There's no issue workflow either. What we do have: change checks. Drift checks. Over datasets we ingest. A diff engine. A change classifier. Schema evolution tracking. A CDC log.	Their core product. Machine-learned checks on freshness, volume, schema, and spread. Alerts to Slack, Teams, email and PagerDuty. Custom SQL monitors on each tier.	◇
Data quality scoring	Each dataset gets a full profile. Completeness, field stats, patterns, column semantics. Plus a relationship graph. Quality metrics, lineage, and compliance flags. On top sits one trust score. Scored per entity type, per domain, across the estate. It breaks down by axis and updates hourly.	They score quality as monitors and issues. Not a profile. Data Insights covers usage and cost instead. We found no per-dataset trust score in their docs.	◆
Lineage	A typed lineage graph. We check access at each hop. It comes from the pipelines that own the data. It's not a stored column-level graph of your warehouse. It doesn't reach into dbt.	Column-level lineage from Pro upward, with impact analysis. It's one of the better ones out there.	◇
PII and PHI classification	129 field classes. They cover PII, PHI, financial data, and secrets. Verdicts are deterministic and reproducible. All 18 HIPAA Safe Harbor identifiers get their own class. Each one cites 45 CFR §164.514(b)(2). Plus source-system fingerprinting across 95 built-ins.	Not part of the product. Their security page says they only touch combined metadata. Row counts. Freshness. Summary stats. Classifying field contents is out of scope by design.	◆
Tokenization and data handling	Deterministic, join-preserving, vault-reversible tokens, applied at ingest. Plus quasi-identifier generalization. Dates become year, decade, or age band. ZIPs become 3 or 4 digits. Phones, SSNs, and emails get partial masking. Each column gets a measured cardinality-reduction score. Turning a token back is privileged and logged.	No tokenization. Their fix: don't read the values. At all. Clean design, for a watch tool.	◆
Audit evidence	SHA-256 hash chain, with Ed25519-signed checkpoints. We chain those checkpoints too. Verification returns clean. Or it names the exact break. Tampering. Insertion. Deletion. Truncation. Try the verifier.	Alert history and issue records. We found no cryptographic tamper proof in their docs.	◆
Agent authorization	Each checked tool call passes a scope ceiling. Plus a consented-tool allowlist and an access tier. A pull-access re-check runs before dispatch too. It fails closed, each time.	Not a thing they sell. They scope access by user and role. In a dashboard.	◆
Break-glass	Scoped, time-boxed emergency access for agents. Admin and step-up gated. IP-allowlisted. Fully audited and auto-revoking.	Not described in their public docs.	◆
GDPR erasure	Crypto-shred of per-subject key material. Plus ISO 27560 consent receipts. The audit chain still checks out. Even after the subject is gone.	They claim GDPR compliance on their security page. They hold metadata, not your rows. So erasure is mostly your warehouse's problem, not theirs.	◆
MCP and agents	More than 200 MCP tools. Across Ontology, Auth, Corpus, and Lighthouse. MCP tool tokens carry scope ceilings. We meter usage per call. We attribute it to the agent.	No AI, agent, or MCP words on their site. Not as of this check. Datadog ships its own MCP server. APM, logs, metrics, monitors, dashboards. We found no Metaplane tools in that mix.	◆
Deployment	Self-hosted, in your own cloud or data center. Or a dedicated single-tenant server we run. Docker images cover Auth, Ontology, Corpus, and Lighthouse. A signed deploy manifest, Guardian checks it. Ed25519 audit-signing keys sit on your instance. In a key directory you control. Rotation fails closed on an unknown key. KMS or HSM custody is a built seam in Auth. It's not a shipped integration yet. Tokenization keys sit in your environment today. Not in a KMS.	SaaS only, on AWS, hosted in the United States. PrivateLink at Enterprise. We found no self-hosted or customer-account option.	◆
Warehouse and connector coverage	Narrow, and we say so. We scan it and profile it. We sort a live PostgreSQL source in place. No rows leave it. Snowflake, BigQuery, Databricks, S3, and Salesforce sit on the roadmap. Still not live. No handler yet. That's PostgreSQL today, not your whole estate.	Broad warehouse coverage. A Snowflake Native App. dbt integration. Job checks too. This is their home turf.	◇
Maturity signals	Auth, Guardian, and Lighthouse are all live in production. Guardian and Lighthouse have run since April 2026. SOC 2 isn't certified yet. We say so on each page.	Founded 2019. Y Combinator. $22.2M raised. Acquired by Datadog in April 2025. SOC 2, GDPR, and HIPAA compliance claimed, reports on request. Around 25 named customers with attributed quotes.	◇
Pricing	Published model. A scoped instant quote. No sales wall.	Free forever at 10 tables and 4 users. Pro is usage-based per monitored table, at 100 tables and 12 users. Enterprise is quote only. No dollar figures published for either.	◈

◆ DataShield leads◇ Metaplane leads◈ comparable

Metaplane claims come from metaplane.dev. Plus Datadog's own blog and docs. Last checked 13 September 2026. We link them below, not from memory.

Three things you get here that you won't get from a data observability platform

Proof an examiner can check

A watch tool writes alerts. Alerts are rows in a database. Rows can get edited. Ours is a hash chain, with signed checkpoints. The verifier names what broke. Tampering. Insertion. Deletion. Truncation. EU AI Act Article 12 reviewers ask about exactly this. So do HIPAA §164.312(b) reviewers. Try it in your browser, no signup.

Authority that changes mid-flight

An analyst hands in notice on a Friday. Their agent is 20 minutes into a 40-minute job. DataShield re-checks the next tool call against current access. It fails closed. A watch layer would tell you Monday. How Auth does it.

A profile that knows what the column means

Row counts and freshness say a column is present. They don't say it holds a medical record number. We sort fields into 129 classes. That includes all 18 HIPAA Safe Harbor identifiers. We give each dataset a trust score. You can track it over time. What Ontology profiles.

Where Metaplane is genuinely stronger

Their whole product is the thing we don't build. Machine-learned anomaly spotting is hard to get right. Freshness, volume, schema, spread, all of it. The hard part isn't the math. It's not drowning the on-call engineer. Metaplane tuned that for years, on mid-market warehouses. The customer quotes read like real ones. A named person. A named number. Not a blended vendor average. Column-level lineage and impact analysis are strong. The free tier is free, not a trial. And Datadog's reach is an edge. No independent data-quality vendor can buy it. That's close to what the founder said. Right when the deal closed.

Here's the push-back. It's about the next two years. Not a knock on the product today. Metaplane hasn't said a word about AI agents. No agent language. No MCP. No per-call authorization, before or after the acquisition. Datadog's own MCP server exposes APM, logs, metrics. And monitors. We could not find Metaplane's data-quality surface in it. Meanwhile, agents are the thing reading your tables now. The question after a broken dashboard used to be "which pipeline?" It's shifting. Now it's "which agent, under whose authority, and can you show me?" A different layer answers that. It's the one we sell. Ask them where it sits on their roadmap. Fair question. They may have a good answer.

Questions worth asking both of us

These are the questions we'd want answered. If we were buying. Ask them on each call. Ours too.

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. The verdict tells deletion apart from truncation and tampering. Run it against a sample chain at /verify. Metaplane: their docs describe alerts and issue logs. We found no tamper-evidence method there. Ask them to show one.

What happens to a revoked agent mid-session?

DataShield re-checks access on each checked tool call. So a pull lands on the very next call. Metaplane doesn't sell agent access control. The honest answer: ask whoever issued the agent its keys. How long does a pulled credential keep working?

How does GDPR erasure interact with the audit trail?

DataShield crypto-shreds per-subject key material. It issues an ISO 27560 consent receipt too. Actor identities in the chain are HMAC-committed. So the proof still checks out. Even once the subject is erased. Metaplane claims GDPR compliance. They hold combined metadata, not your rows. So erasure is mostly a question for your warehouse. Ask both of us for the method. Not the policy page.

Is DataShield a data observability tool? Do we drop Metaplane?

No, and no. We have no anomaly checks. No monitors on your tables, no freshness SLAs. There's no issue queue either. Silent breakage is your problem? Buy a watch tool. Run us on the datasets agents read instead. There, the job is classification, authorization, and proof. Plenty of teams sensibly run both.

Metaplane is owned by Datadog now. Does that change anything?

It might. The April 2025 announcement said the experience stays the same "for now." Careful phrasing. Datadog's plan: fold data watching into one platform. Infrastructure, application, and data, together. Good news if you already run Datadog. Less good if you bought Metaplane for its free tier. Or for its focus. Ask them what the free tier looks like in 2027. We're small and independent, which cuts both ways. Our design-partner terms include source escrow. A small vendor isn't a single point of failure. Not that way.

Does DataShield have SOC 2?

Not yet. We won't fake it. Metaplane claims SOC 2, GDPR, and HIPAA. Reports on request, on their security page. We publish a threat model instead. Plus a verifier anyone can run. Plus an honest limitations list. Auth is live now. Guardian and Lighthouse have run since April 2026. Details on the security page.

Metaplane's hero: "Be the first to know about data incidents" and "an end-to-end data observability platform that catches silent data quality issues before they impact your business." - metaplane.dev, 13 Sep 2026
Datadog acquired Metaplane on 23 April 2025. The product continues standalone as Metaplane by Datadog, and "all features, support, and services continue uninterrupted." - Metaplane blog, 23 Apr 2025
Datadog's stated plan: unify infrastructure, application and data monitoring, building on Data Jobs Monitoring and Data Streams Monitoring. - Datadog blog, 23 Apr 2025
Tiers as published: Free (10 monitored tables, 4 users, 3 custom SQL monitors), Pro (usage-based, 100 tables, 12 users, column-level lineage), Enterprise (custom, SSO, PrivateLink). No dollar figures for Pro or Enterprise. - metaplane.dev/pricing, 13 Sep 2026
Metaplane "only accesses aggregate metadata, such as row counts, freshness, and summary statistics." SaaS on AWS, hosted in the United States. SOC 2, GDPR and HIPAA compliance claimed. - metaplane.dev/security, 13 Sep 2026
Datadog's MCP server documents coverage of APM, logs, metrics, monitors, dashboards and security signals. No Metaplane data-quality tools appear in its documented surface. - Datadog docs, 13 Sep 2026

Other head-to-heads

Observability

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 →