Built for GitHub Copilot

Agents build it.
Your people sign it off.

Lockstep turns the GitHub Copilot you already have into a governed delivery flow, from business intent to production. Copilot does the work; a named reviewer approves every phase.

Works with
GitHub CopilotVS CodeGitHubMCPYour Copilot models
01 intent

New intent

What should the agents build?

Morning, Priya!

Describe a business goal. Lockstep turns it into a requirements spec for you to approve.

Describe the outcome you need…
lifecycle-activity

Lifecycle activity

Requirements Agent42s
Spec Approved
Architect Agent1m
Build Agent38s
Code Review
Evaluation Agent2m
Policy Check6s
Release Agent51s
Observability Agent12s
02 level-1-plan

Level-1 plan

5/5 steps
human-gate

Promote to production?

All five evaluation thresholds pass. Rollout starts on 10% of new applications with the runbook attached.

adlc/05-release.md

artifacts

Artifacts

Governance

03 bolts

Bolts

3/5 done

  1. {"tests":14}

    14 tests pass. Field extraction is 97% accurate on the labelled sample.

  2. {"lists":2}

    screen.sanctionsparallel

    Consolidated sanctions lists, daily refresh.

    screen.pepparallel

    PEP list via the approved provider.

  3. {"tests":11}

    Every flag cites the documents it compared.

  4. {"mcp":"kyc-system","mode":"dry-run"}

  5. {"cases":400}

04 code-review
agents/kyc/risk-policy.ts+2 −1
export const riskPolicy = {
autoApprove: ['low', 'medium'],
autoApprove: ['low'], // NFR-2
approverRole: 'kyc-analyst',
};
faq

Code and artifacts stay in your repos. Model calls go through your existing Copilot plan.

Lifecycle

From intent to impact,
with a human gate after every phase

AI drives planning, decomposition, generation and execution. People validate decisions, provide oversight and confirm business value before the next phase can start. Open any phase to read the skill file that runs it.

01 discover & define
02 architect & design
03 build & orchestrate
04 evaluate & validate
05 release & operate
06 observe & evolve

Continuous feedback loop: the improvement backlog from phase 06 becomes the next intent. Learn, improve, repeat.

ADLC Foundation

Built on what you already run,
governed by your enablement team

foundation
  • One model connectionUses your Copilot models. No second key, nothing on laptops.
  • Named-role gatesPull request reviews routed by CODEOWNERS to the right team.
  • Separation of dutiesGitHub stops authors approving their own phase.
  • Fails closedBranch protection blocks the next phase until approval.
  • Centrally managed packsYour enablement team approves every skill pack update.
  • Client-hosted registryPacks live in your GitHub or artifact registry.
  • Tools / MCPWorks with Copilot’s MCP registry and allow-list.
  • Cost per phaseSee the Copilot credits each phase used.
  • Traceability & AuditEvery approval attributed, versioned and exportable.
traceability & audit
Gate and agent events for intent 0192, newest first.
16:42Priya Nairrelease.approved
16:39Agenteval.scored
16:31Tomás Reyesdiff.accepted
16:30Agentpolicy.checked
16:18Agentbolt.completed
15:57Agentmcp.called
15:40Lena Fischerplan.approved
15:36Agentplan.proposed
15:12Lena Fischeradr.signed
14:58Priya Nairspec.approved
14:51Agentspec.revised
14:44Priya Nairspec.changes_requested
14:30Agentspec.drafted
14:21Priya Nairintent.filed
14 rows
Use cases

Applicable across
agentic product development

data-engineering

Data Engineering

  • Data product design and requirements
  • Pipeline generation and orchestration
  • Automated DQ and testing
  • Metadata and knowledge graph integration
business-automation

Business Automation

  • Process discovery and documentation
  • Agentic workflow design
  • System integrations via APIs and MCP
  • Continuous monitoring and optimisation
genai-rag

GenAI / RAG Solutions

  • Knowledge design and ingestion
  • RAG plus agent architecture
  • Evaluation, guardrails and safety
  • Enterprise deployment and scaling
software-engineering

Software Engineering

  • AI-assisted development
  • Code generation and review
  • Automated testing
  • Release management and observability
Pricing

Simple plans,
on top of the Copilot you already have

Lockstep uses your Copilot models, so there is no separate model bill from us. Approvers who only review and sign are always free.

dev

Dev

For individual developers running the lifecycle in their own repos.

$9per month
Get Dev
  • @lockstep phases in Copilot Chat
  • Generic lifecycle pack
  • Personal approvals
  • Packs from the Lockstep registry
  • Community support and docs
team · most teams start here

Team

For engineering teams that need governed delivery, not just faster code.

$19per developer / month
Start with Team
  • Everything in Dev
  • Named-role gates via CODEOWNERS and branch protection
  • One industry pack included
  • Team admin and seat management
  • Audit export
  • Email support
enterprise

Enterprise

For regulated enterprises that host everything in their own environment.

Contact usannual licence, volume pricing
Contact sales
  • Everything in Team
  • Client-hosted pack registry with approved updates
  • All relevant industry packs, plus custom packs
  • Risk-based extra approvers
  • SSO, central policy and offline licensing
  • Pilot, onboarding and support SLA

Prices in USD, excluding taxes. Requires a paid GitHub Copilot plan (Business or Enterprise for Team and Enterprise). Phases use your Copilot models and count towards your Copilot usage, and Lockstep shows the credits each phase used.

FAQ

Things teams ask
before a pilot

No. Lockstep runs inside GitHub Copilot in VS Code and uses the models your Copilot plan already provides, including your own models if you have configured them in Copilot. There is no second model connection and no model keys on developer laptops.