Terms of Service

Last updated: October 1, 2026

1. Acceptance of Terms

By installing the dollama CLI, connecting to the dollama relay (api.dollama.net), using the network to serve inference requests, or contributing compute as a node operator, you agree to these Terms of Service ("Terms"). If you don't agree, don't use the network.

These Terms should be read together with the Privacy Policy, which describes what happens to your data.

2. Description of the Service

dollama connects computers into a shared pool for LLM inference: your own machines, machines in groups you join, and, if you opt in, a network of volunteers' machines. It is an experimental developer tool. There is no company operating dollama as a commercial service; the hosted relay and this website are operated by an individual, Lia Pool. The relay matches requests to machines and carries the request and the response between them; it does not run inference itself. It holds each request for a few minutes while it is being served and keeps no prompt or response content after that (see the Privacy Policy §5).

3. Eligibility

You must be able to form a legally binding contract to use the network — generally, you must be at least the age of majority in your jurisdiction, or have the permission of a parent or guardian.

4. Using the Network for Inference

• Best-effort, no SLA. Inference is provided by volunteers on a best-effort basis. There is no guaranteed uptime, latency, throughput, or model quality. The network may be slow, unavailable, or route your request to a machine with different performance characteristics than another user's.
• No warranty on model output. Model responses are generated by third-party open-weight models running on volunteer hardware. dollama does not review, curate, or guarantee the accuracy, safety, or appropriateness of any model output. Treat model output the way you would from any LLM provider — verify before you rely on it.
• Fair-use tracking. Your usage (token counts, request volume) is tracked to enforce fair-use rate limits and to compute your contribution balance. See §6 on the balance/priority system.
• Plaintext exposure. In every mode, your prompts and responses can be read by the relay and by the machine that processes them (see the Privacy Policy §3 for which machines, per mode). Requests are not yet encrypted between your machine and the machine that processes them. Don't send secrets, credentials, or anything you wouldn't send to a third-party cloud API.

5. Contributing Compute (Node Operators)

If you choose to run dollama in a mode that shares your hardware with the network, the following applies:

• Voluntary, revocable. Contributing is entirely optional and can be stopped at any time by switching modes or stopping dollama. There is no obligation to keep your node online.
• Your hardware, others' requests. Other users' inference requests may run on your machine while you're contributing. You are responsible for ensuring your hardware, network connection, and local laws permit this.
• Your hardware first. Your own requests are routed to your own node ahead of other people's, and are not counted against your balance. That is prioritisation, not a guarantee: a request already running on your node, or a model that has to load, can still delay yours.
• No payment, no financialization. There is no monetary compensation for contributing compute. Contribution is tracked as a running balance used only to prioritize your own requests when the network is busy — it is not a token, cryptocurrency, or tradeable asset, and cannot be bought, sold, or transferred.
• Bandwidth & resource usage. Serving requests consumes your bandwidth, GPU/CPU cycles, and electricity. You are responsible for these costs.
• No inspection guarantee. We cannot verify what you do with content that passes through your node while serving requests. By contributing, you agree not to log, store, inspect, or otherwise misuse content you process on behalf of other users.

6. Acceptable Use

You agree not to use dollama — as a requester or as a node operator — to:

• Generate or distribute content that is illegal in the jurisdiction where you or the content's recipients are located.
• Attempt to disrupt, overload, or degrade the network's availability for other users (e.g. denial-of-service behavior).
• Attempt to circumvent rate limits, fair-use tracking, or the balance/priority system through automation or account manipulation.
• Attempt to gain unauthorized access to another user's account, API key, node, or content.
• Submit sensitive personal data, credentials, or regulated data (e.g. health, financial, or legally privileged information) in any mode, including Private, given the plaintext exposure described in §4 and the Privacy Policy.
• As a node operator, run modified software that misrepresents your node's capabilities, benchmark results, or identity to the relay.

We may suspend or revoke access (API keys, node registrations) for accounts or nodes that violate this section, without prior notice, on a best-effort/community-moderation basis appropriate to a volunteer project.

7. No Warranty

8. Limitation of Liability

To the maximum extent permitted by applicable law, in no event shall dollama's maintainers, contributors, or node operators be liable for any indirect, incidental, special, consequential, or punitive damages, or any loss of data, profits, or goodwill, arising out of or related to your use of (or inability to use) the network — including damages resulting from model output, network unavailability, or another node operator's or requester's conduct.

This is a volunteer, community-run project. There is no commercial entity to indemnify losses, and no service-level agreement backs any use of the network.

9. Intellectual Property

The dollama source code has not been published yet. When it is, it will carry an open-source licence, and that licence will govern the code. Until then, you may install and run the released binaries to use dollama under these Terms. You retain all rights to the prompts, files, and content you submit through the network — dollama and its contributors claim no ownership over your content and do not use it for any purpose beyond routing and generating the response you requested.

10. Termination

You may stop using dollama, revoke your API key, or stop contributing compute at any time. We may suspend or terminate access for violations of §6 (Acceptable Use) or where required to protect the network's availability or other users. For ordinary inference requests the relay does not keep prompt or response content long term: each request sits in the relay's cache only for the short periods listed in the Privacy Policy §5 (minutes, not days) and then expires on its own, so none of that content remains to delete on termination. Some things are stored: account data, usage records, evaluation results you submitted, and evaluation transcripts if you opted in to assisted evaluation. These are handled as described in the Privacy Policy §11.

11. Changes to These Terms

As dollama is under active development, these Terms may change to reflect new features (for example, group sharing, new privacy modes, or client-to-client encryption). We'll update the "Last updated" date above when we do. Continued use of the network after a change constitutes acceptance of the updated Terms.

12. Governing Law

These Terms are provided as a general framework for a volunteer project and do not designate a specific governing jurisdiction. If you require a specific governing-law clause for your use case (e.g. organizational deployment), please raise it at the contact address below.

13. Contact

dollama is operated by Lia Pool. Questions about these Terms should go to privacy@dollama.net. Security reports should go to security@dollama.net.

Read the Privacy Policy →