Weave FAQ: AI ROI, Engineering Output, Integrations and Data Access

What is Weave?

Weave connects engineering output, code quality, AI usage, and spending. Engineering leaders use it to understand what their teams ship and how AI tooling relates to that work.

The measurement products answer different questions:

• Engineering Intelligence covers output, review health, quality signals, delivery measures, and benchmarks.
• Token Intelligence covers AI usage, subscription and usage costs, attribution, and the relationship between spending and output.
• Code Intelligence provides the Silk 1 model that scores engineering output and explains its estimates.

How does Weave measure engineering output?

Weave evaluates the source changes in a PR and estimates how long an expert engineer would need to complete that work. Silk 1 considers intent, complexity, risk, and dependencies across files. A score of 3.0 represents three estimated expert hours of output.

This is a standardized estimate of work, not a timesheet or an automatic estimate of hours saved by AI. It should be interpreted alongside the underlying changes, review context, and quality outcomes. Read the measurement methodology for the definitions and limitations.

How does Weave connect AI spending to output?

Token Intelligence combines subscription configuration with billing and usage telemetry. It relates available spending and AI activity to engineering output, with attribution granularity depending on the connected tools.

The public Weave Index calculates cost per output hour using engineers with both attributed spending and merged output in the same month. It excludes ignored PRs and spending that cannot be attributed to a person. The result describes observed cost efficiency. A financial ROI claim additionally needs a credible baseline, a value for incremental benefits, and total incremental costs. It does not follow from an output ratio alone.

How is AI contribution attributed?

Weave distinguishes native tool API evidence, inferred Git co-authorship, and estimated attribution from activity windows. These signals have different strengths. Native data, commit metadata, and an estimate should not be treated as equivalent proof that a tool produced a particular change.

A PR can include several tools and human edits. Review the attribution method and coverage before comparing tools or interpreting a change in AI-assisted output. Token Intelligence explains the available signal hierarchy.

Which AI tools and cost breakdowns are supported?

Capabilities vary by integration. Token Intelligence's public per-user cost coverage includes Amazon Q, Claude Code, Cline, Cursor, Firebender, Sourcegraph, and Windsurf. Tools without per-user tracking use organization-level costs. Native contribution APIs are available for tools such as Cursor and Copilot; other attribution methods use the available metadata or activity signals.

For a particular rollout, verify the tool, plan, repository host, historical coverage, and required permissions with Weave. A logo or a general AI coverage claim does not establish that every tool supports identical usage, cost, and output fields. The Claude Code, Cursor, and Codex guides explain the questions to check.

Does Weave access source code?

Code Intelligence analyzes source code changes in pull requests, including the full diff and relationships across files. Describing the product as accessing metadata only would be inaccurate.

The data and permissions required depend on the connected product and integration. Review the actual authorization request, privacy policy, security information, and subprocessor information during evaluation. Request the current details for retention, model providers, access controls, and data residency that apply to your configuration. This FAQ does not replace those documents or your agreement.

What security documentation is available?

Weave publishes security policies and controls and maintains a trust center. Use the current trust materials and the review process for your organization to verify certification scope, deployment requirements, and contractual terms. Contact the team through a demo request for documentation needed for procurement.

Which quality and delivery signals can I review?

Engineering Intelligence reports review quality and cycles, reverts, code turnover, and deployment measures. These signals help assess whether changes are accepted, require repair, or create delivery problems. They also provide context for output and spending comparisons.

Define an observation window for post-merge outcomes. Recently merged work may not yet reveal defects or maintenance cost. A model's output estimate is not a substitute for reviewing those results.

What do the benchmarks represent?

The Weave Index and Weave Research report aggregate observations from the data described in each methodology. The population, period, denominator, and exclusions matter when comparing a team with a published figure.

These are observational samples. They do not establish a universal adoption target or prove that an AI tool caused a productivity change. Compare similar work and teams, and use your own historical baseline alongside external data.

Can Weave compare multiple AI tools?

Token Intelligence provides a cross-tool view of spending, usage, and engineering output where data is connected. A fair comparison still depends on comparable coverage, attributed costs, task types, and quality outcomes. If several assistants contribute to the same PR, avoid assigning the full value of the PR to each assistant independently.

For a renewal decision, ask to trace a representative usage record and cost to the corresponding engineering evidence. Record gaps rather than treating disconnected sources as zero activity.

How should we evaluate Weave?

Choose one concrete decision, such as a tool renewal or an adoption rollout. Use a representative repository and confirm the identity mapping, permissions, billing sources, historical coverage, and output definitions. Review several PRs with the engineers who know the work, then compare costs and outcomes for a consistent cohort and time window.

The time needed for meaningful findings depends on the integrations, available history, and time required to observe quality outcomes. Discuss those inputs when booking a demo, and use the AI ROI measurement methodology to structure the evaluation.

Where can I find pricing?

See current Weave pricing and enterprise information. Confirm feature coverage and contractual requirements for your organization with the team.

Product definitions and references reviewed September 22, 2026.