Skip to content
Alayra Systems

Kinetic IDE · sovereign AI IDE

In development

Intelligence without surveillance.

A local-first AI development environment. Kinetic indexes your codebase on your own machine, works through an agent you can watch, and asks before it changes files or runs commands.

Get early accesskinetic-ide.com
Kinetic IDE with an orders API open; the Kinetic AI panel shows a finished task — snippets retrieved from the local index, files edited and tests run

Three gates of trust

An agent that earns its place in production.

01

Local RAG engine

A Rust indexer chunks and embeds your workspace on your machine — vector plus BM25 hybrid retrieval, Tree-sitter symbols, re-indexed in the background on every save.

02

Blueprint consent

Before a multi-file change, the agent writes a Blueprint: the plan, the files it will touch and a risk badge. You approve, reject or revise.

03

Sentry gate

Any command outside the approved plan stops inline in chat with a risk score — low to critical — and waits for you. Every decision is logged.

Blueprint · Mission · Trace

Consent to the plan before a single file changes.

Code mode begins with a Blueprint you can read. Once approved, a Mission tracks the steps and a Trace records what actually happened — all saved as files in your workspace, so every agent turn leaves an audit trail you can browse.
  • Approve · Reject · Revise
  • Risk badge on every plan
  • Pulse Timeline of every turn
Kinetic IDE showing a Blueprint awaiting approval, with the files it will touch

Sentry

Nothing runs without your knowledge.

When the agent wants to run a command the plan doesn’t cover, Kinetic stops and shows it to you. Trusted commands can be allow-listed; everything else is scored, approved or rejected, and kept in a searchable log.
  • Risk levels: low, medium, high, critical
  • Trusted-command allow-list
  • Searchable Sentry log
Kinetic IDE pausing on a shell command for approval in the Sentry gate

Mission Command

The state of your work, at a glance.

A home for the active mission, the pulse timeline, the model and key pool in use, workspace health and smart suggestions — so you always know what the agent is doing and why.
Kinetic IDE Mission Command home tab

Any model

Your keys, any provider — or nothing leaves at all.

Use Kinetic AI Auto, add your own key for 25+ providers, or run fully local models through Ollama or LM Studio. A context meter shows exactly how much of the model’s window each task uses.
Kinetic IDE Add Custom AI panel configured for a local Ollama model

Four modes

The mode you pick is a ceiling.

A classifier chooses the lane that actually runs, and a policy decides which tools each lane may use — so a question never becomes a refactor.

Ask

Read-only. Questions about your code, answered from the local index.

Plan

Writes plans and artifacts only — never touches your source.

Code

The full agent loop, with the Blueprint and Sentry gates in front of every change.

Vision

Code mode plus images, URLs and browser tooling.

The privacy model, precisely

What stays, and what travels.

Always local
Your index, embeddings and vectors; credentials in OS-backed secure storage; Blueprint, Sentry and Trace logs.
With a cloud model
Your prompt and only the snippets a task needs go to the provider you chose. Alayra does not see or store them.
With a local model
Nothing leaves your machine.
What Alayra keeps
Only what it takes to run your account: your email and basic details, token counts and usage, and credit and billing records. Never your code or your prompts.

25+ providers, plus local

Bring the model you trust.

  • OpenAI
  • Anthropic
  • Google Gemini
  • Vertex AI
  • Azure OpenAI
  • AWS Bedrock
  • Groq
  • OpenRouter
  • Mistral
  • Together
  • Cohere
  • Fireworks
  • Perplexity
  • DeepInfra
  • Replicate
  • DeepSeek
  • xAI
  • Hugging Face
  • Kimi
  • NVIDIA NIM
  • Cerebras
  • SambaNova
  • AI21
  • Novita
  • Lambda
  • Cloudflare AI
  • Ollama
  • LM Studio

On the horizon

Squad Mode and team indexing.

A leader, architect, coder and QA agent working as one team — and shared, remote and sharded indexes that run on your own infrastructure. Both are planned, after the core gates stabilise.

See the roadmap

FAQ

Questions, answered

Does my code leave my machine?

Your codebase is indexed locally, and embeddings and vectors stay on your machine. With a cloud model, your prompt and only the snippets a task needs go to the provider you choose; Alayra does not see or store them. With a local model through Ollama or LM Studio, nothing leaves at all.

Which models can I use?

Kinetic AI Auto, your own keys for 25+ providers — from OpenAI and Anthropic to Groq, Mistral, Bedrock and Vertex — or fully local models. Keys are kept in the IDE’s secure storage.

Is Kinetic IDE built on VS Code?

Kinetic IDE is built on Code-OSS, the open-source core of VS Code, with a Rust engine for orchestration, indexing and retrieval.

When can I use it?

Kinetic IDE is in active development for Windows, macOS and Linux. Join early access and we will reach out as builds open.

Early access

Every Alayra product is in active development. Join early access for Nexus Cloud, Kinetic IDE and Kinetic AIDE — or self-host Nexus today.

Get early accesshello@alayrasystems.com