AI platform
Your organisation's AI platform. Not an API key.
Define your agents, skills, tools, and knowledge once. They work the same in chat, in Slack, in the terminal, through the API, and in workflows — under your guardrails, your spend limits, and one audit trail.
Six building blocks, one governed platform
The idea
One definition. Every surface.
Most AI platforms give you an endpoint and leave the rest to you. Every team then rebuilds the same agent for each place it is needed, and the audit trail breaks at every boundary.
An agent you define here is available to your whole team the moment you save it. The same agent answers in the dashboard, responds to a Slack mention, runs in a developer’s terminal, reviews a pull request from a webhook, and executes a scheduled workflow overnight.
You do not rebuild it per surface, and you do not lose the audit trail when it moves between them.
What the platform gives you
Four things your organisation defines centrally, and everything else follows from them.
Agents
Specialists for code, planning, security, review, and debugging in every organisation. Add your own with explicit tool permissions, or customise a platform agent without forking it.
Agents
Skills
Import from skills.sh or GitHub, or write your own. Write a skill once and assign it to any agent. Built-in skills cover security, accessibility, and privacy.
Skills
Workflows
Visual node-based automation with webhook and schedule triggers. Pause for human approval. Run agents over a backlog with durable orchestrations.
Workflows
Governance
Guardrail presets, model allowlists, spend caps per user and per interface, filter policies, and an audit log nobody can edit.
Governance
Knowledge
Ground answers in your documents
Create collections, upload documents with automatic chunking, and search them semantically. Agents read from them directly, so answers cite your material rather than the model’s recollection.
- Automatic chunking with configurable overlap
- Metadata filtering on semantic results
- Query one collection or every collection at once
- Multiple embedding models
Autonomous engineering
Agents that finish the job
Long-running work runs in its own container. It clones the repository onto a branch, works through the tasks, and opens a pull request. Repository audits, dependency reviews, multi-step refactors.
Build on it
Or just use the API
Everything above sits on an API you can call directly. There is an OpenAI-compatible endpoint, so pointing an existing application at Quant is a base URL change rather than a rewrite.
OpenAI-compatible chat completions, embeddings, and models
Buffered REST, server-sent events, WebSocket, or durable async
- Images, video, and documents as input
- Embeddings with a choice of models
Optional context compaction for long agent conversations
28 of 30 models run with Australian-resident inference. On QuantGov, you see only those 28.
What are the key benefits of edge deployment?
• Lower latency - Content served from nearby locations
• Better reliability - No single point of failure
• Reduced bandwidth - Less traffic to origin servers
• Smaller attack surface - Origin servers stay out of public view
What people build
From a single assistant to a fleet that runs unattended.
AI-powered search
Add semantic search and grounded chat to any website with our managed search product. Crawling, indexing, and retrieval included.
- Semantic and keyword search
- Automatic crawling
- Grounded chat
Retrieval over your documents
Vector collections, embeddings, and chat combined so answers cite your material instead of the model guessing.
- Automatic chunking
- Source attribution
- Metadata filtering
Unattended engineering work
Scheduled audits, pull request reviews, and multi-step refactors that run in a container and open a pull request.
- Runs on a schedule
- Opens pull requests
- Fully logged
Conversational assistants
Multi-turn chat with session memory and streaming, in your product or in Slack where your team already works.
- Session memory
- Real-time streaming
- Slack native
Event-driven automation
A webhook starts a workflow, agents do the work, and the result lands wherever you need it. No polling, no glue code.
- Webhook triggered
- Human approval steps
- Versioned definitions
AI a risk committee will sign off
Mandatory guardrails, model allowlists, spend ceilings, and an audit log that cannot be edited after the fact.
- Mandatory presets
- Pre-request spend checks
- Immutable audit
Start with the platform, not the plumbing
Agents, skills, tools, and knowledge — governed from day one.