Beta benchmark targets
  • AZURE OPENAI · EASTUS2 · P50 218 MS
  • AMAZON BEDROCK · US-EAST-1 · P50 236 MS
  • GEMINI 4 ARGON · GOOGLE CLOUD · 1M CTX · P50 208 MS
  • GUARANI VERIFY · DETERMINISTIC PR GATE
  • MOBILE CI · REAL ANDROID / iOS DEVICES
  • SEMANTIC CACHE HIT · 46.2%

Private beta · Trial workspaces on request

One explained verdict for every release.

Guarani AI reviews every pull request, routes and meters every model call, and runs your mobile journeys on real devices, then turns the findings into one release decision you can trace.

Guarani AINVIDIA Inception Program

PROGRAM MEMBER · 2026Guarani LLC is a member of NVIDIA Inception, the NVIDIA program for startups.

guarani · acme/core-apiVRF-1842 · CLEAREDSample run
  1. guarani verify --repo acme/core-api --profile release
  2. passed: Indexed 1,284 files, semantic graph ready3.1 s
  3. passed: Architecture rules: 42 checked, 0 violations4.8 s
  4. advisory: API contract: 1 additive change on GET /v2/orders2.2 s
  5. passed: SAST: 0 new findings, no secret added in the diff3.6 s
  6. passed: Performance budget: p95 within budget (+1.8%)6.4 s
  7. passed: Generated 12 unit and 3 integration tests, passing9.7 s
  8. passed: Mesh: 214 model calls, p95 532 ms, $0.38live
  9. passed: Mobile journeys: sign-in, checkout on 3 real devices41.0 s
  10. CLEARED for release, 1 advisory. Report posted to pull request #1842 as a check.

Run summary

Files indexed
1,284
Architecture rules
42 checked, 0 violations
API contract
1 additive change, v3 compatible
SAST
0 new findings, 0 secrets
p95 budget (450 ms)
within, +1.8%
Tests generated
12 unit, 3 integration
Model calls
214, p95 532 ms, $0.38
Device journeys
sign-in, checkout on 3 devices

Pipelineacme/core-api

  1. Indexqueued3.1 s
  2. Architecturequeued4.8 s
  3. API contractqueued2.2 s
  4. SASTqueued3.6 s
  5. Performancequeued6.4 s
  6. Testsqueued9.7 s
  7. Meshqueuedlive
  8. Mobilequeued41.0 s

LLM latency by providerp50 p95

  • Azure OpenAI218 / 540 ms
  • Amazon Bedrock236 / 575 ms
  • Gemini 4 Argon208 / 520 ms
  • GPU open-weight172 / 390 ms

Running

CLEARED for release

1 advisory, 0 failures. Check posted to pull request #1842.

Gateway overhead budget
<18ms
Max context window
1M+
Cloud and GPU inference routes
4
Android and iOS device runners
Real

One release gate for code, models and mobile.

Each module installs where the work already happens: Verify on GitHub and GitLab pull requests, Mesh in front of your model providers, Mobile CI inside GitHub Actions or GitLab CI. Use one module or all three: versioned policy rules combine their findings into one verdict.

Sample run

Verify builds a semantic graph of your codebase and reviews every pull request against it: architecture rules, API contracts, performance budgets and static security analysis (SAST). Findings are posted to the pull request before it merges.

  • Semantic pull request review across files, services and ownership boundaries
  • Static application security testing (SAST): unsafe data flows from untrusted input to sensitive calls, hard-coded secrets and unsafe API usage
  • Early detection of performance regressions against latency and memory budgets
  • API contract validation for REST, GraphQL and gRPC, with breaking-change gates
  • Automatic generation of unit and integration tests for uncovered paths

Verdicts come from versioned policy rules applied to the findings: the same findings and the same policy always produce the same verdict.

Explore Guarani Verify
Pull request reviewSample review
#1842orders: add fulfillment ETA
acme/core-apifeat/fulfillment-eta into main6 files changed
Verify, architecture agentPerformanceservices/orders/eta.ts:48

estimateEta() calls the inventory service once per line item, inside the loop. On large orders this adds sequential calls and pushes GET /v2/orders past its 450 ms p95 budget.

Batch the lookups into one getAvailability() call to stay under budget.

Suggested change
- for (const item of order.items) {-   const stock = await inventory.get(item.sku);-   etas.push(estimate(item, stock));- }+ const skus = order.items.map((item) => item.sku);+ const stock = await inventory.getAvailability(skus);+ const etas = order.items.map((item) => estimate(item, stock[item.sku]));
Verify, architecture agentContractopenapi/orders.yaml

Additive field fulfillment_eta on GET /v2/orders. Existing clients are unaffected: compatible with contract v3.

Verify, security agentSecurityservices/orders/eta.ts

SAST: no new data flow from request input to a query, shell or file call, and no secret added in this diff.

Generated 4 unit tests and 1 integration test for estimateEta(), all passing.

One OpenAI-compatible endpoint for every model your teams use. Mesh routes each request by latency, cost, context length and data residency, fails over across regions and providers, and meters spend per team.

  • Unified OpenAI-compatible API in front of Azure OpenAI, Amazon Bedrock, Google Gemini models on Gemini Enterprise Agent Platform and open-weight models on dedicated GPUs
  • Dynamic routing by latency, cost, context length and data residency
  • Multi-region failover with health checks and retries
  • Semantic context cache for repeated and near-identical prompts, scoped to your tenant
  • Per-team cost tracking, budgets and alerts
Explore Guarani Mesh
Routing policy
Sample data
Traffic share, latency and spend by provider over 24 hours, for the selected policy
ProviderSharep50p95Spend / 24 h
Azure OpenAIeastus2, swedencentral38%218 ms540 ms$77.33
Amazon Bedrockus-east-1, us-west-231%236 ms575 ms$47.32
Gemini 4 Argon (Google Cloud)us-central1, 1M context21%208 ms520 ms$50.75
GPU (open-weight)dedicated endpoint10%172 ms390 ms$7.00
Weighted p50
217 ms
Weighted p95
532 ms
Spend, 24 h
$182.40

Spend by team, daily budget

  • payments$61.1093% of daily budget
  • search$44.2074% of daily budget
  • platform$38.7077% of daily budget
  • mobile$24.9062% of daily budget
  • data$13.5045% of daily budget

Run UI and regression suites on real Android and iOS devices from the pipelines you already have. Every result carries its video, device logs and network traces.

  • Native GitHub Actions and GitLab CI integration
  • Real device pools allocated per run
  • Parallel UI and non-regression suites (Appium, Espresso, XCUITest, Maestro)
  • Flaky test detection with retry and quarantine
  • Video, device logs and network traces on every result
Explore Guarani Mobile CI
checkout suite on 3 real devices
Sample run
Result of each journey on each device
Devicesign-insearchcheckoutpayment sheet
Pixel 10passed12.8 spassed13.9 spassed21.0 spassed9.7 s
iPhone 17passed12.1 spassed13.2 spassed20.4 spassed9.9 s
Galaxy S25passed13.3 spassed14.4 sHOLD, retry passed22.6 spassed10.2 s
12 resultsvideodevice logsnetwork traces

How a pull request moves through Guarani AI.

Every change runs as a traced pipeline across managed cloud regions, GPU inference and real device pools. Customer code is processed in isolated tenants and never used to train models.

  1. Webhook received

    A pull request on GitHub or GitLab notifies the control plane at api.guarani-llc.com/v1. The policy engine resolves the profile for the repository.

  2. Code indexed

    The orchestrator updates the semantic graph and the vector index for the changed files, inside the customer's isolated tenant.

  3. Review agents dispatched

    Architecture, contract, performance and security (SAST) agents run on analysis workers: containers on AKS, ECS/EKS and Cloud Run.

  4. Mesh routes model calls

    Every agent call goes through Guarani Mesh. When Azure OpenAI degrades in eastus2, the router marks the call REROUTED to swedencentral and the trace continues.

  5. Mobile suites run

    The affected journeys run on real Android and iOS devices. Video, logs and network traces are stored with the run.

  6. CLEARED for release

    Versioned policy rules mark the strip CLEARED, the verdict is posted to the pull request as a check, and every step is written to an audit trail.

Source and CI

  • GitHub
  • GitLab
  • GitHub Actions
  • GitLab CI

Control plane

  • APIapi.guarani-llc.com/v1
  • Orchestrator
  • Policy engine
  • Tenant isolation

Analysis workers

  • Review agentsarchitecture, contract, performance, security
  • ContainersAKS, ECS/EKS, Cloud Run

Device cloud

  • Real device poolsAndroid and iOS
  • Artifactsvideo, logs, traces

Guarani Mesh

  • Router
  • Semantic cache
  • Budget guard
  • Telemetry

Model providers

  • Azure OpenAIeastus2 degradedswedencentral
  • Amazon Bedrock
  • Gemini Enterprise Agent PlatformGemini 4 Argon, 1M context
  • Open-weight modelsdedicated GPUs

Data

  • Object storage
  • Postgres
  • Vector index
  • Telemetry warehouse

Data handling

  • Encryption in transit and at rest.
  • Processing in US or EU regions chosen per workspace.
  • No training on customer code, prompts or outputs.
  • Retention you configure for repository content and artifacts.

Deployment options

  • Multi-tenant cloud with isolated tenants, the default for every plan.
  • Dedicated VPC on Azure, AWS or Google Cloud.
  • Isolated GPU inference endpoints for open-weight models.

Works from your terminal, your pipeline and your code.

Install the CLI, add one step to your workflow or call the REST API. Typed SDKs for TypeScript and Python wrap the same endpoints.

terminal
# Install the CLI (Node.js 20 or later)
npm i -g @guarani-ai/cli

# Authenticate with your workspace access token
guarani login

# Create guarani.yml for this repository
guarani init

# Run the release profile and print a JSON report
guarani verify --repo acme/core-api --profile release --json

The packages and endpoints shown are the v1 beta interface.

Private beta clients are published on npm and PyPI. Remote execution requires a GUARANI_TOKEN from a provisioned Private Beta workspace.

Read the docs

Three plans with published quotas.

Every plan starts with a 14-day free trial. Prices in USD, billed monthly, taxes excluded.

Starter

For a team putting its first repositories under review.

$99/ month

  • Up to 5 repositories
  • 5M analysis tokens per month
  • Automated pull request reviews
  • 100 mobile CI test minutes per month
Start 14-Day Free Trial

Enterprise

For dedicated infrastructure, isolated GPU inference and audit evidence.

Custom

  • Dedicated VPC deployment on Azure, AWS or Google Cloud
  • Isolated GPU inference endpoints
  • Audit trails and evidence exports for SOC 2 and ISO 27001 programs
  • Up to 99.99% uptime SLA
  • Dedicated support engineer
Contact Enterprise Sales

Compare plans and read the FAQ

Team

The engineers building Guarani AI.

About Guarani AI
Sylvain Rey

Sylvain Rey

Founder & Principal Systems Architect

Designs and builds the Guarani AI platform: the control plane, the review agents, the Mesh gateway and the CI integrations.

Thomas Moreau

Thomas Moreau

CTO

Leads GPU inference and mobile CI infrastructure.

Verify your next release.

RLS-NEXTyour-org/your-repoAWAITING CLEARANCE

Start Free Trial

Private beta clients are published on npm and PyPI. Remote execution requires a GUARANI_TOKEN from a provisioned Private Beta workspace.