Enterprise DNA

Instead of Weights & Biases Weave

Stop renting Weights & Biases Weave. Own your evaluation records.

We build your version around the answers your business must get right, bring your selected Weave records across and run the agreed review process with you. Your team keeps the evidence behind every decision.

The vendor’s AWS Marketplace offer lists a 12-month W&B Weave 10GB commitment at US$25,000, with additional usage charges. This is one contract offer, not the entry price. Free and Pro plans also exist.

Get your version built See it working
  • check_circleYour data in a database you own
  • check_circleOpen source, no lock-in
  • check_circleBuilt and run by Enterprise DNA

The Weave bill

Compare your actual contract with the work you need

Weave’s pricing address now leads to CoreWeave Forge. The current plans still include Weave, and the vendor’s Marketplace listing carries a separate annual offer. Check which contract and capabilities your team actually uses.

paid

An annual commitment

The vendor lists the W&B Weave 10GB commitment at US$25,000 for 12 months on AWS Marketplace. Private agreements and additional usage can change the bill.

storage

Logged information

The current pricing page measures Weave ingestion by information received and stored, including recorded inputs and outputs.

folder

Stored evidence

The vendor describes storage separately from ingestion. Check how long you retain answer records and what your contract includes.

compare_arrows

Evaluation work

Weave brings test examples, scoring and recorded results together. The owned base focuses on saved answers and the acceptance checks you agree.

timeline

Live tracing

Weave also captures application activity for investigation. Include the live capture your team depends on when scoping your own version.

support_agent

Hosting and support

The current plans distinguish support and enterprise capabilities. Compare those responsibilities with a version built and operated for your business.

Your own evaluation desk

The expectation, the answer and the decision in one place

Coverage

Missing answers stay visible

Register the cases your business expects a version to answer, then check the complete set.

  • check Keep the expected answer separately from the answer being tested.
  • check Group cases by policy, formatting or fallback behaviour.
  • check Count a missing answer as a failed check.
  • check Apply your own response-time limit to each case.

Comparison

See exactly what got worse

Compare two versions against the same acceptance cases.

  • check Show each case that passed before and now fails.
  • check Separate answer failures from response-time failures.
  • check Count failures by the business group you chose.
  • check Refuse comparisons when the case definitions differ.

Evidence

Keep the basis of each review

Save the rules and answers behind a result before anyone signs it off.

  • check Preserve the inputs used for each saved evaluation.
  • check Flag old results when expectations change.
  • check Record the actual reviewer and their evidence.
  • check Require a different reviewer before accepting a passing result.

Weekly work

Turn failures into owned follow-ups

Give the team a concrete review pack and a list of work to resolve.

  • check Assign each follow-up an owner and due date.
  • check Record the evidence that closes an action.
  • check Produce branded decision packets and data handling reviews.
  • check Draft the weekly review from run attention, open actions and handling dates.

Make it yours

Design your own Weights & Biases Weave replacement

Everything starts from the open-source base. Then it gets your name, the parts you run, your rules and your data from Weights & Biases Weave. Try it here, and we build the real one.

2. What it runs

3. Make it fit

4. Your colour

Your business · your Weights & Biases Weave replacement Your version
YB Your business

auto_awesome

Answered from your own database

Build my version

We build it from the open-source base, bring your Weights & Biases Weave data across and run it for you.

What it looks like

Ask it like you would ask a colleague

Real answers from the demo data that ships with Evaluation Runs for Claude Code. It runs in Claude Code inside the Claude app, so it feels like any chat you already use.

forum

Evaluation Runs for Claude Code

Connected to your own database

Which answers passed before and now fail?

order-status

Cohort
structured
Baseline
pass
Candidate
pass
Change
unchanged
Reason
passed

refund-route

Cohort
policy
Baseline
pass
Candidate
fail
Change
regression
Reason
latency exceeded

returns-window

Cohort
policy
Baseline
pass
Candidate
fail
Change
regression
Reason
contains failed

unknown-order

Cohort
fallback
Baseline
pass
Candidate
fail
Change
regression
Reason
missing answer

Ask it anything

Ten questions the free version answers today

These use the fictional demo and the shipped checks. Weave also supports configurable analysis. The difference is that you own this version and its rules.

chat_bubble Which answers passed before and now fail?
chat_bubble Did the candidate leave any expected answers out?
chat_bubble Which correct answers take longer than our limit?
chat_bubble Which business group has the most failed checks?
chat_bubble Which saved decisions refer to changed expectations?
chat_bubble What exact answers and rules supported an earlier result?
chat_bubble Who reviewed this result and what evidence did they give?
chat_bubble Which overdue follow-ups need an owner to act?
chat_bubble Which data access and retention reviews are overdue?
chat_bubble What needs the team lead’s attention this week?

Switching

How to switch from Weights & Biases Weave

A move usually involves choosing which call records to keep, mapping them to your acceptance cases, then testing the imported history before you retire Weave.

  1. 1

    Export selected call records

    Choose the Weave call records you want to bring across, then export them as JSON, JSONL or CSV. Check the export filters and columns so you know which records are included. You do not have to bring over all of your history. Minimise personal information in the export and agree who can access it.

  2. 2

    Map records to cases

    Each exported call needs to map to an existing case in your acceptance suite. Decide which Weave field identifies that case and how your selected records correspond to the acceptance checks you want to keep. The example uses a user-defined case key. Weave does not supply that field automatically, so check your own records before importing.

  3. 3

    Test, then import the records

    Ask for a test import first. It checks the selected export and its mapping before writing records to the database. Review the test results, then proceed with the import when the mapping and source count are right. The import preserves the original call records and records the mapping used, so you can trace what came across.

  4. 4

    Review import findings

    Check what the import reports before relying on the imported history. It rejects calls with unknown cases, duplicate calls, mixed source projects, invalid timings or a source count that does not match. Resolve any problems in the export or mapping, then test again. Reconcile the source count so you know the imported records match the records you selected.

  5. 5

    Compare, then switch off Weave

    Keep Weave and the new database in use side by side for a short period. Run your saved checks and compare results with an earlier run using the same cases. Confirm that the records and evidence you need are available, and have a reviewer assess the results. Once your team is comfortable with the new process, stop using Weave for this work.

construction

Where the real effort goes

The main work is usually deciding which history to bring across, mapping your own fields to acceptance cases, and agreeing which checks and rules matter. The import can preserve call records, but existing Weave scores remain source evidence rather than newly verified results. If you have Enterprise DNA install and run your version, they handle the export mapping, test import and count reconciliation with you.

Not ready for a call?

Get your Weights & Biases Weave switch plan by email

We will send you the step-by-step plan for moving your business off Weights & Biases Weave, what the move really involves, and how we would do it for you.

One email with your plan. No spam, and you can unsubscribe any time.

Three ways to own it

Do it yourself, customise it, or have us run it

Open source

Do it yourself

Free, MIT licence

  1. 1 Clone the repo and run npm install, then npm run demo.
  2. 2 Open it in Claude Code, Codex, OpenCode or Cursor.
  3. 3 Set up your cases and map your Weave export, then test the import.
  4. 4 Run your checks, compare versions and keep a private backup.
Get it on GitHub

Built around your business

We customise it

Setup fee, then a retainer

  • check Your expected answers, review rules and business groups.
  • check Your selected Weave history mapped and reconciled.
  • check A web view and team access built for how you work.
  • check A different stack or additional scoring where your version needs it.
Get your version built

Omni by Enterprise DNA

We run it for you

Setup fee, then a monthly retainer

  • check Installed and operated through Omni by Enterprise DNA.
  • check Capture connections scoped around the systems you already use.
  • check Backups, access and data retention responsibilities agreed.
  • check A review process with clear owners and evidence for each decision.
Book a call with Sam

Questions

Instead of Weights & Biases Weave, answered

Does everyone using Weave pay the annual figure above? expand_more

No. The US$25,000 figure is the vendor’s listed annual W&B Weave 10GB commitment on AWS Marketplace. Its current pricing page also offers Free and Pro plans, with Enterprise terms agreed separately. Compare your own bill and usage before choosing a replacement.

Can our existing answers and history come across? expand_more

Weave documents exports of call records. We map the selected answers to your agreed acceptance cases, reconcile the source count and do a test import. The free version retains the original call records and reruns its own checks. Existing vendor scores are kept as source evidence rather than treated as newly verified results.

Will this cover our live capture and specialist checks? expand_more

The free version checks saved answers for exact matches, required wording, valid structured output and response times. Enterprise DNA scopes live capture, more advanced scoring, team access and a web view for your custom version. We agree that scope against the capabilities you use today.

Does a passing result mean the assistant is safe to release? expand_more

A pass means the stated checks passed. It is evidence for a human decision. Missing answers and recorded errors fail, changed expectations make old evidence stale, and acceptance requires a different reviewer. Your business still decides whether the checks cover the intended use.

Why use this when Weave already has an open-source toolkit? expand_more

Weave is a broader tracing and evaluation toolkit. This is a focused record of your acceptance cases, saved answers, comparisons and review evidence. You can run the free version yourself, or have Enterprise DNA build and operate your version for a setup fee and retainer. The point is ownership and fit for your process.

Put your evaluation bill beside the work it supports

Spend 30 minutes with Sam looking at your Weave contract, a sample export and the decisions your team needs to make. We map the version your business would own and what Enterprise DNA would build and run.

Book a 30-minute call with Sam Get the free version

Evaluation Runs for Claude Code is free and open source, published by Enterprise DNA.

Done paying for Weights & Biases Weave?

Get your version built