Connect GitHub
Sign up through GitHub and authorize the read-only connection for your organization.
How it works
From engineering evidence to better decisions.
Start with the questions that matter. Connect your selected engineering sources, inspect the evidence, and choose a practical next move.
The hosted trial starts with read-only GitHub access. Explore other connection and deployment options.
A focused way to begin
Your engineering evidence.
A clearer working view.
Start with GitHub Cloud for self-service setup and a selected set of repositories. For other providers, private networks, legacy systems, or project-management context, explore the integration and advanced setup options. Reports become available as collection and analysis complete.
Choose your scope
Select the repositories you want to understand. Keep the scope relevant to your team's questions.
Explore the evidence
Review available reports, collection status, and findings as your engineering evidence is processed.
Works with your engineering toolchain
Connect the work.
Improve the practice.
Explore GitHub, GitLab, Bitbucket, and self-hosted Git, plus advanced setup for legacy repositories and project-management context. Build an AI-forward engineering practice around the tools your team uses.
Explore integrations
Choose your sources and agree the setup for your environment.
Run Binomial on your infrastructure
Buy the software. Get the source code.
Receive the Binomial source code and Dockerized software, and deploy it on your own infrastructure. Keep using your last entitled version perpetually. Your purchase includes 12 months of released product features and enhancements.
Explore Self-Hosted
Optional Annual Maintenance & Support from Year 2.
Hosted trial · GitHub example
A connection with
clear boundaries.
Authorize GitHub Cloud and choose the repositories to include. Collection runs asynchronously; reports reflect the available persisted evidence.
Review access and evidence boundaries
Authorize GitHub Cloud
A read-only connection to your engineering source.
Select your repositories
Choose the scope that matches your team's questions.
Collect and analyze
Asynchronous processing. Check freshness and completeness.
Explore the available view
Follow findings into their evidence and decide what to investigate.
GitHub observations support engineering analysis. Direct AI-tool usage requires separate authoritative evidence.
Source and scope
Connected engineering evidence.
Example organization · Selected repository scope
Repositories in scope
3 selected- platform-corePlatformIncluded
- api-serviceProductIncluded
- data-pipelineDataIncluded
Connection context
- Provider
- GitHub Cloud
- Access
- Read-only
- Collection
- Asynchronous
- Before interpreting
- Check freshness and completeness
Authorize connectionSelect repositoriesCollect and analyze
Reports become available as collection and analysis complete.
- 01
See your source
The connected provider and organization stay visible.
- 02
Check the scope
Selected repositories define the evidence included.
- 03
Review collection state
Use readiness and freshness to interpret what is available.
Observed
Pull requests, reviews, and repository context.
Interpreted
A modeled explanation with confidence and limitations.
Discussed
Team context and a focused next action.
An observed change does not establish that AI caused it.
See the reasoning, not just the result
One finding. Its evidence. A practical next move.
Every finding separates what was observed from what the evidence suggests, then keeps confidence and limitations beside the recommendation.
Four example walkthroughs
Understand the evidence boundariesWhich engineers are using AI effectively?
Start with how the work is approached and validated. GitHub evidence can support an authorized coaching conversation; establishing AI usage and impact also requires appropriate provider evidence and team context.
Focused changes make the approach easier to review.
- Pull requests
- Review discussions
- Validation notes
- Observed evidence
- In this example, an engineer’s focused changes include clear acceptance checks and validation notes. Review discussions retain the reasoning behind the work.
- Modeled interpretation
- The approach may offer a useful coaching example. These observations alone do not establish whether AI was used or whether it improved the outcome.
- Confidence
- Directional. The working practices are visible; AI effectiveness has not been established.
- Limitations
- Direct AI-tool telemetry and work outside the selected GitHub scope are absent. This is not an employee ranking or productivity score.
- Next action
- Discuss the approach with the engineer through authorized coaching, and identify a practice the team can test.
Which practices should spread across teams?
Look for practices worth trying in another setting. Keep the task, team, and review context attached as you compare patterns.
Smaller changes are moving through review faster.
- Change size
- Review events
- Follow-up work
- Observed evidence
- In this example window, smaller pull requests require fewer review cycles than larger changes in the same selected repositories.
- Modeled interpretation
- Focused changes and earlier validation may help reviewers. The pattern suggests a practice to investigate, rather than proof that AI caused the difference.
- Confidence
- Moderate within the example window. The pattern is consistent in the selected work.
- Limitations
- Task complexity, team conventions, and work outside GitHub may explain part of the difference.
- Next action
- Pilot focused change sets and explicit acceptance checks on comparable work, then review the next window with the team.
Which repositories are ready for AI-assisted work?
Examine whether changes can be understood, tested, and reviewed reliably. Prioritize the conditions that help the team validate assisted work.
Validation evidence is uneven around a busy module.
- Selected code
- Workflow signals
- Module documentation
- Observed evidence
- The example repository has sparse documentation and inconsistent test signals around a frequently changed module.
- Modeled interpretation
- Focused tests and clearer module guidance may make assisted changes easier to evaluate and maintain.
- Confidence
- Directional. Additional codebase and team context is needed.
- Limitations
- GitHub workflow evidence does not capture every local test. Collection freshness and completeness affect what is visible.
- Next action
- Strengthen tests and documentation around the selected module, then reassess validation and review patterns.
Which investments will improve velocity without increasing debt?
Turn an observed constraint into a bounded investment hypothesis. Agree on the cost, scope, and outcome to examine before expanding the intervention.
Strengthen validation before expanding generation.
- Review patterns
- Follow-up fixes
- Validation context
- Observed evidence
- In this example, follow-up fixes cluster around larger changes with limited validation evidence.
- Modeled interpretation
- Targeted validation work may address a delivery constraint. Its cost and expected benefit need confirmation with the team.
- Confidence
- A working hypothesis. Financial assumptions and a later comparison window are needed.
- Limitations
- Follow-up fixes may reflect planned iteration or changing requirements. The example cannot guarantee faster delivery or lower technical debt.
- Next action
- Choose one module for a focused test investment. Compare similar changes, review effort, and follow-up work before deciding whether to expand.
One connected decision package
Six outputs connect people, practices, codebases, and investment.
Together, they show where coaching, codebase investment, and operating changes are most likely to improve engineering leverage.
Individual AI effectiveness profiles
Evidence-linked strengths, friction, and coaching priorities.
Team capability map
Practices to spread, review bottlenecks, and enablement needs.
Repository readiness assessment
Codebases evaluated for safe, effective AI-assisted work.
Technical-debt roadmap
Debt prioritized by its effect on delivery and AI leverage.
AI engineering operating model
Workflow standards, review expectations, and governance guidance.
Measurement baseline
Current effectiveness, delivery, quality, and risk signals.
Scope and availability depend on the connected evidence and any agreed expert review. AI-use claims require authoritative provider evidence. Individual profiles support authorized coaching and development, with confidence and limitations attached.
Tools and guides
Work through the decision.
Explore a scenario, understand the evidence model, or get help defining a focused assessment.
Interactive tool
AI Engineering Value Model
Model capacity, rollout costs, training time, and payback using your assumptions. All outputs are modeled estimates.
Explore the value modelMethodology
How to read the evidence
Understand source coverage, observed evidence, modeled interpretations, authorized context, and follow-up measurement.
Read the methodologyOptional expert support
A bounded readiness assessment
Agree on the engineering question, selected sources, reviewers, and priorities before expanding the scope.
Explore the assessmentBefore you begin
A few practical questions.
Which source-control systems can I connect?
The hosted trial starts with GitHub Cloud. GitLab, Bitbucket, and generic Git have additional connection paths; private and on-premises Git require an agreed network and access setup. Legacy systems such as SVN (Subversion), CVS, Perforce, and SCCS, plus project-management sources, are advanced setup topics. Confirm compatibility and available evidence for your environment before purchase. Explore integrations and setup options.
Will reports appear immediately?
Collection and analysis run asynchronously. Availability depends on your selected scope and collection state. Review freshness and completeness before interpreting a report.
Can repository activity measure AI-tool usage?
Not by itself. Commits and pull requests show how code changed, but they do not reliably show whether someone used AI. Measuring actual usage requires data from the AI tools themselves. Without that data, we cannot confirm AI usage.
Can I get help applying the findings?
Yes. A readiness assessment and capability program provide optional expert support.
Start with your engineering evidence
A clearer picture.
A better next move.
Connect your engineering sources and start understanding the patterns behind your outcomes. Try the hosted version with GitHub, or get the source code and deploy Binomial in your own environment.