Everything an orchestrator should be.
One engine that handles local and remote models, built-in tools, persistent agents, and team workspaces. No subscription needed to run the engine itself, and local mode stays free forever.
The capability map sits under everything on this page.
Every feature here reads from the same place. Soriku benchmarks each model you connect, on your own hardware, across 16 categories. Each score comes with its spread and the number of tests behind it. What was never measured says so, and what got old is flagged. Routing, verification, gap analysis and team workspaces all read from that map.
- One benchmark run per model, 16 categories, on your hardware
- Every score shows its spread and how many tests it came from
- Re-run whenever you want, and stale measurements get marked
Orchestration
Smart routing
Capability-scored routing per category. Override per prompt or pin globally.
Multi-model verify
Run a prompt through two or three models, return the best or merged answer.
Capability benchmark
Measure your models on your machine. 16 categories, re-runnable, yours.
Stack optimiser
Soriku spots gaps in your model coverage and suggests fills.
Request triage
Soriku reads each prompt and decides whether one model can handle it or whether it needs planning out across several. You never pick the mode yourself.
Synthesis you can trace
When more than one model works on a task, Soriku merges the results, takes a consensus, or keeps the strongest answer, and marks which part came from which model.
Models & providers
Ollama (local)
First-class. Auto-discovery of installed models, RAM tuning, mute/reactivate flow.
Claude, OpenAI, Mistral, Groq
Bring your own keys. Per-provider daily budget caps.
OpenAI-compatible endpoint
Drop-in replacement at /api/v1/chat/completions for any SDK or tool.
RAM-aware loading
Soriku knows your hardware and won't load three 8B models at once.
No rate-limit tax
Local models have no requests-per-day or tokens-per-minute caps. Limits only ever apply to the cloud calls you choose to make.
Watches for new models
Pull a new model into Ollama and Soriku picks it up within half a minute, scores it, and starts routing to it. No restart, no manual setup.
Agents & tools
Agent personas
Persistent identities with their own memory, tone, specialisation.
Pilot mode
Multi-step tool calling. Read files, run shell, fetch web, generate documents, either supervised or fully automatic.
Plan mode
Multi-worker planning with conductor + editor. Cost-balanced or fully local.
MCP integration
Soriku speaks Model Context Protocol. Use it in Cursor, VS Code, or the terminal.
Built-in voice
Speech-to-text runs locally on Whisper, with no metered audio-seconds. Optional Mistral Voxtral fallback when you want the cloud.
Learning memory
An agent picks up how you work: your writing style, the stack you're in, the trade-offs you tend to make. It carries that into the next conversation and lets older habits fade.
Knowledge links
Point an agent at files or folders and they stay in its context. When a linked file changes, the agent notices.
Built-in tools
Read and write files, run shell, fetch the web, generate a PDF, a Word doc, or a spreadsheet. All of it runs on your own machine.
Teams & deployment
Simezu SSO
One Simezu account signs you in across the whole Atypisch ecosystem, with shared groups and shared billing.
Shared agents
Roll out an agent persona to your team. Everyone gets the same setup.
Team capability map
Aggregate scorecard across your team's hardware. Nobody re-benchmarks.
Self-host or hosted
Run on your hardware, or let Soriku Cloud handle it.
Rate limits and quotas
Hosted workspaces get per-tenant rate limits and per-plan quotas, so one heavy run doesn't starve the rest of the team.
Knowledge & context
Project memory
Every project keeps its own memory. Decisions, facts, and the way you like things done stay scoped to that project and come back when you return to it.
Recall by meaning
Soriku searches a project's past conversations by meaning rather than keywords and brings the relevant pieces back into the prompt.
Grounded in your sources
Point Soriku at your codebase or your docs and answers come back grounded in them, with the source attached so you can check it.
Reads your documents
Hand it a PDF or a long document and Soriku parses it into something a model can reason over.
Artifacts you keep
When an answer contains a document, a code file, or an export, Soriku lifts it out as a real file you can download.
Trust & monitoring
Every answer has provenance
You always see which model answered and why it was chosen. Nothing gets routed behind your back.
Feedback that lands
Mark an answer good or bad and it feeds back into how routing and your agents behave the next time.
Stays within your machine
Soriku reads memory pressure as it runs and adapts what it loads, so a large model doesn't bring everything to a halt.
What it looks like at your desk.
