← Vetro Attest · Free Art. 12/14 checker · Free EU AI Act tools

Vetro Attest vs. LangSmith

*Engineering comparison, not legal advice. Check LangSmith's own documentation for its current features, plans and hosting options.*

LangSmith is LangChain's platform for building and operating LLM applications. Vetro Attest is a small, single-purpose service that countersigns an agent log so it can serve as evidence. They solve different problems, and they combine well.

What LangSmith does well

LangSmith is widely used by teams building agents, with or without the LangChain framework. Its strengths:

• Tracing each step of a chain or agent: prompts, model calls, tool calls, retrievals, with latency, tokens and errors;
• Evaluation: datasets, experiments, human and automated feedback to compare versions before shipping;
• Monitoring production traffic with dashboards and alerts;
• Tight integration with LangChain and LangGraph, plus SDKs and OpenTelemetry support for other stacks.

If your team is iterating on an agent, it gives you visibility and a feedback loop that are hard to build yourself.

What it is not designed to give you

LangSmith is a tool for the team that builds the agent. Evidence is aimed at people who do not have to trust that team.

• The data lives in a platform operated by you or your vendor. Traces can be deleted according to retention settings or by users with the right permissions. That is appropriate for an engineering tool, but it means an export is your word about what happened.
• No independent, timestamped signature. There is no signed statement from an outside party that a specific trace existed, unaltered, at a given time.
• No hash chain that travels with the file. A trace export does not let the recipient detect that a step was removed or reordered after the fact.
• No mapping onto EU AI Act logging duties. Art. 12 asks providers to make high-risk systems record events that allow traceability; Art. 19 and 26 cover keeping those logs. A monitoring dashboard is not organised around those questions.

What Vetro Attest adds

Vetro Attest takes the trace you already have: an OTLP/JSON export using OpenTelemetry GenAI conventions, or a native Vetro ledger. It rebuilds it into a hash-chained ledger and verifies it for free. For 29 USD per ledger, one-off:

• a full EU AI Act report (Markdown and HTML) mapping the run onto logging obligations;
• an ed25519 countersignature by this service over the digest, event count and timestamp, checkable by anyone using our published public key and a free verification endpoint.

What it does not do: it does not attest that what the agent did was correct or true, and the report is not legal advice or a certification. Paid ledgers are deleted after 7 days; the report and attestation are emailed to you to keep.

When to use both

1. Build, evaluate and monitor with LangSmith.
2. For runs with legal or commercial weight, such as a high-risk decision, a customer complaint or an incident, export the trace in OTLP/JSON if your setup supports it, and get it countersigned soon after the run. The earlier the countersignature, the stronger the "it looked like this then" claim.
3. Keep the attestation JSON with the trace in your retention storage (ideally WORM).

If you export traces in another format, the free checker will still show you where your logging stands against Art. 12 and 14.

Try it

• Free EU AI Act Art. 12/14 checker, in your browser.
• Verify a trace free, then get the report and countersignature for 29 USD.
• Background reading: audit trail for AI agents.

See a sample ledger verified in one click, check your own logs free or get the countersigned EU AI Act report.