Your organization’s memory, ready to use
A ready-to-use workspace for teams: chat sessions, a live calendar and inbox, the organization's memory and its apps. Connect the places information already lives, then ask, act and build — with the sources attached.
// app.miriel.ai — by keyboard or voice
Ask anything, with sources
Chat sessions over everything your organization has connected. Answers carry inline citations keyed to the passages they came from, and a session remembers what it has already learned.
The day, already in context
A live inbox and calendar alongside the memory, so a meeting arrives with what was promised last time, and a message can be drafted from what the organization already knows.
A memory you can correct
Inspect every item, see its chunks and its source, browse the extracted entities as a graph — and forget anything that should not be there.
Apps from one sentence
Describe a process and Miriel compiles a readable plan: the steps, what triggers them, and exactly which data the app may touch. Test it before turning it on.
Work you can see
Boards, goals, dashboards and forms built on the same memory, so the numbers and the status come from the sources rather than from someone updating a slide.
No learning curve
A one-box assistant, by keyboard or voice, serves anyone who wants no learning curve — and a command palette for everyone who does.
A browsable connector library for the systems your team runs, plus files, URLs, pasted text and HTTP endpoints for everything else. Queryable in seconds, and it stays in sync.
// six sources syncing into one queryable memory
Every person, project, ticket, document and source becomes an entity, linked to the others it appears with. Browse it as a graph in the workspace; queries walk it, not just the index.
// entities and relationships extracted at ingest, settling into one graph
Inline citations keyed to the passages they came from, real stage-by-stage timings, and an honest error instead of a plausible answer when a query cannot run.
POST /api/v2/query/timedprojectleeloonum_results10modelclaude-sonnet-5// one timed query: route → search → entities → graph → distill → compose
Permissions travel with the data at the token level. Switch a grant off and the next answer is composed without it — no re-index, no second copy.
// the same question twice: grant-56 (finance/*) on, then off
Works with the leading AI model providers and with models you host yourself.
- Switch models within a single conversation, even between providers, and the context carries over.
- The memory belongs to you, not to any model: the same memory serves every provider, and can be shared across people, teams and AI tools.
- Per-user permissions apply to everything Miriel stores, searches and returns, so answers never cross access boundaries.
- Heightened privacy mode keeps content out of logs and routes AI calls to a private provider that does not retain data.
The workspace is one way in. The same secure memory sits behind your AI tools and your own software too.
Inside your AI tools
Miriel hosts MCP servers, so Claude, ChatGPT, GitHub Copilot, Cursor and custom agents can learn into and query the same memory as a tool.
Inside your software
One API and one credential to learn, query and build apps — Python, C#, TypeScript or REST — so Miriel can sit behind products you already run.
GET STARTED
Open the workspace.
Connect a source, ask a question, and see the sources behind the answer.
// no spam, just a beta waitlist