Amari Code
Agentic coding for practical software engineering
Amari Code plans, edits, and verifies real changes across your codebase — running entirely inside your infrastructure, so your source code and IP never leave your control.
Capabilities
An engineering agent, not an autocomplete
Repo-aware reasoning
Reasons over your full repository — modules, dependencies, and history — to make changes that fit your architecture.
Multi-step refactoring
Plans and executes multi-file refactors step by step, keeping APIs and behavior consistent across the codebase.
Test generation
Writes unit, integration, and regression tests for the code it touches, raising coverage where it actually matters.
Tool use in CI
Invokes your build, test, and lint tools directly in CI pipelines — verifying its own work before a human reviews it.
How it works
A sovereign agentic workflow
Every step runs within your perimeter, on models you govern.
# Amari Code: agentic engineering workflow
task = repo.issue(#1247) # "Migrate billing to new pricing API"
plan = amari.plan(task, context: repo.full_index)
plan.review(required: true) # human-in-the-loop gate
amari.edit(plan.files) # multi-file, repo-aware changes
amari.generate_tests(coverage: affected_modules)
ci.run([:lint, :typecheck, :test]) # tool use in your own CI
amari.fix(ci.failures) until ci.green?
pr = amari.open_pull_request(
summary: plan.rationale,
evidence: ci.report
) # ready for your team's reviewFor developers
Integrate Amari Code into your stack
API references, CI integration guides, and deployment options for VPC, on-premises, and Kanda Vault environments.
Ready to deploy sovereign AI?
Run Kervalt models in your own cloud, on your own terms.