The frontend to Bittensor
Bittensor has world-class infrastructure and few places where everyday people use it. Ditto is an AI memory app that people and their agents use every day, and it already pays subnets for the services it runs on. Plug your subnet into Ditto and turn Ditto's users into revenue for your subnet.
Subnet 22Desearch: search
Subnet 75Hippius: file storage
Subnet 64Chutes: modelsWhat Ditto runs on today
Ditto pays subnets in production for the services it uses, and runs its own subnet to improve its memory.
Ditto's own subnet
Subnet 118
Miners compete to build the best memory harness, scored on DittoBench. The winning work runs in Ditto: temporal search and keyword search came from miners.
Desearch
Subnet 22
Web and X search in Ditto run through Desearch.
Hippius
Subnet 75
Users store the files they upload and the files Ditto generates in their own Hippius bucket.
Chutes
Subnet 64
Chutes is one of the model catalogues in the Ditto app, alongside OpenRouter, OpenAI, Anthropic and Google.
Minos
Subnet 107
Minos ships a Docker image with Ditto pre-installed, so their miners start from shared best practice instead of from zero.
What Ditto gives a subnet
Most subnets have strong infrastructure but little paid demand from everyday users. Ditto brings that demand.
New revenue from Ditto's users
Ditto pays the subnets it uses in production. When your service sits behind Ditto, its users become paying demand for your subnet.
Every agent that connects to Ditto
Claude, Cursor, Codex, OpenClaw and Hermes read and write Ditto's memory over MCP, so work routed through Ditto reaches agents on every platform.
A place in Ditto's plans
Ditto sells plans to people, teams and businesses. Subnet services can be packaged into those plans, so users pay for Ditto and your subnet is part of what they get.
Users from outside Bittensor
People sign up to Ditto with no wallet and no card, so your service reaches users the network doesn't reach today.
Memory around your output
What your subnet returns becomes part of the user's memory, so later answers and agents build on it.
Ditto inside your own stack
Like Minos, a subnet can ship Ditto to its own miners and users, with a shared memory already set up.
Where subnets fit in Ditto
Ditto already has a place for each of these services. A subnet that does one of them well can plug in behind it.
-
Models and inference
The model catalogue in the Ditto app, where users pick a model for each chat.
-
Search and data
Research answers, where Ditto searches memory first and then the web.
-
Storage
Files, uploads and generated artifacts.
-
Agents and code
Agent jobs Ditto runs for users after they approve them.
-
Evaluation
Scoring Ditto's harness and the models behind it on DittoBench.
-
Specialist services
Anything a user or business asks Ditto to do that a subnet already does best.
How a subnet plugs in
- 1
Get in touch
Tell us what your subnet does and who uses it today.
- 2
Find its place in Ditto
We agree which part of Ditto your service sits behind.
- 3
Integrate it
We connect your subnet to Ditto and test it on real use.
- 4
Package it
Your service goes live for Ditto users as part of their plan.
Questions from subnets
What does 'the frontend to Bittensor' mean?
Ditto is the app everyday people and their agents use. Subnets sit behind it and provide services such as search, storage and models, and Ditto pays them for what its users use.
Which subnets does Ditto use today?
Ditto runs its own Subnet 118 and uses Desearch (Subnet 22) for search, Hippius (Subnet 75) for file storage and Chutes (Subnet 64) as one of its model catalogues. Minos (Subnet 107) ships Ditto pre-installed.
Do Ditto users need a wallet or crypto?
No. They sign up free with no wallet and no card. Bittensor runs underneath Ditto.
How does a subnet get paid?
Ditto pays the subnets it uses in production, so its users' activity becomes revenue for your subnet. Get in touch to talk through how that would work for yours.
What is Subnet 118?
Ditto's own subnet. Miners compete to build the best memory harness, scored on DittoBench, and the winning work goes into Ditto.
Is Ditto open source?
Ditto's memory harness is open source, and so are its integrations: the MCP server, the CLI, the desktop app, and the Claude Code, Hermes and OpenClaw plugins.
Is user memory used to train AI models?
No. Memory in Ditto is not used to train AI models.
Plug your subnet into Ditto.
Tell us what your subnet does and we'll find its place in Ditto.
Read-only by default