NewCost control: a token allowance a month for each model choice — when one runs out, the cheaper one answers.See how it works
memtro
Comparison

Memtro vs Dust

A platform for building and sharing assistants, versus a memory your existing clients reach with nothing to configure.

This comparison describes Dust as we understand it at the time of writing from its public materials. Products change; check the current details before deciding.

Dust is a platform for building and sharing AI assistants on company data. You connect sources, assemble assistants with instructions and tools, and publish them to colleagues; it is open source, with a hosted offering, and is strong on letting non-engineers compose agents.

Memtro is a memory rather than a builder. There is nothing to assemble: point an existing client at the endpoint and it gets the organisation's memory, tools over connected systems that read unless you grant an action, routing and jobs. The unit of work is a question, not an assistant somebody configured first.

Where they differ

MemtroDust
Unit of workA request to one endpoint, with memory and tools already attachedAn assistant you configure, then share
SetupConnect a system, point a client at the endpointBuild the assistant that will use the data
MemoryA shared graph that every client and job writes to and reads fromPer-assistant context and conversation
TimeBitemporal: read memory as of any date, or walk what changedCurrent state
Where it runsYour editors, chat UIs, agents, jobs and MCP clientsPrimarily the Dust workspace
ModelsYour own keys, routed by difficultySeveral providers, through the platform
SourceHosted productOpen source, with a hosted option

Honestly, where Dust is stronger

If you want many purpose-built assistants, each with its own instructions and audience, Dust's builder is a better fit than an endpoint — and being open source matters if you need to run it yourself or read the code before trusting it.

When to choose which

Choose Dust when the thing you want to create is assistants: distinct, configured, shared with particular teams, each doing a defined job.

Choose Memtro when you want one memory of the company that everything already in use can reach, and you would rather not configure an assistant before someone can ask a question. If you build on Dust, Memtro can still sit underneath as the memory and connector layer.