08 AwwO workspaces
AwwO is ClawHunt’s workspace for organizing AI agents into teams, available as a web app and a Mac client. This clause covers the hosted AwwO service. If you run the open-source edition yourself, this clause does not cover data on your systems, but requests you send to ClawHunt services such as LLM Gate are covered by this policy.
What we store. The email address and display name from your ClawHunt sign-in and a link to that account, and, for accounts created before ClawHunt sign-in, a hashed password; your workspace memberships, roles and invites; your plan and credits, and any requests for more usage, which we may follow up with paid-plan offers; your appearance settings; and the content you create in AwwO: canvases, agent and node settings, prompts, conversations, run records and outputs, files that agents produce, knowledge you save, and tasks you hand to direct execution, together with the permissions you grant during them. For each model call we also record the model used, token counts, timings, an estimated cost and any error code, for billing, quotas and reliability. Keys you add yourself are stored encrypted.
AwwO keeps you signed in with its own essential cookies. Your browser or the Mac app also keeps copies of canvases you open, unsaved changes and pending requests. Signing out or deleting a canvas does not remove them; in a browser, clearing the site’s data does.
Who processes it. When you message the canvas assistant, plan a canvas, run a node or hand off a task, the prompt and the context it needs are sent to a model service: the one you chose, or ClawHunt’s default model for planning, for routing assistant messages and whenever you have not chosen one. These services are ClawHunt’s model gateway (LLM Gate), which passes requests to third-party model providers and resellers; Amazon Bedrock; the DeepSeek API, for DeepSeek models only (see below); RunningHub, which passes image and video requests to each model’s provider; your own LLM Gate key, if you connect one; and, where enabled for your workspace, TypeSafe for planning decisions. Third-party providers handle requests under their own terms, which we do not control. The hosted service runs on UCloud cloud servers in Tokyo, Japan, and its traffic passes through Cloudflare’s network; Cloudflare and our servers process technical request data such as IP addresses to deliver and protect the service. Model providers may process requests in other countries.
DeepSeek models. AwwO lists DeepSeek’s models under their DeepSeek names and runs them through the DeepSeek API of Hangzhou DeepSeek Artificial Intelligence Co., Ltd. A request goes to DeepSeek only when a DeepSeek model is selected for it, by you or by the canvas assistant when you ask it to change a model: AwwO’s planning, its routing of assistant messages and its own default model never use DeepSeek, and AwwO never moves a request to DeepSeek by itself. The prompt, the context it needs and any pictures attached to it are then sent to DeepSeek, which processes them in the People’s Republic of China under its own terms and Chinese law. DeepSeek’s terms do not rule out using what it receives to improve its models, and copies DeepSeek keeps follow its terms. Do not select a DeepSeek model for personal data or other information you do not want processed in China or used to improve AI models.
AwwO Local. AwwO 本地版 (AwwO Local) runs AwwO on your own computer and keeps what it stores there. If you connect it to your hosted AwwO account with a connection key made in AwwO’s settings (设置 → 本机版连接, AwwO Local connection), the model calls it makes through that connection, AwwO 云端 (AwwO Cloud), go to the hosted AwwO in Tokyo, which passes each one to Amazon Bedrock and counts it as a call of the workspace the key belongs to. For these calls the hosted service records what it records for any model call (the model, token counts, timings, an estimated cost and any error code), but not their prompts or answers. It stores each connection key as a hash and its first few characters, with its name, its workspace and when it was made, last used, expires or was revoked. A key lasts 90 days, and you can revoke it in the same settings at any time.
Who can see it. Members of a workspace can see its canvases and runs according to their role, and one another’s names and email addresses. Authorized ClawHunt staff can view account, workspace and run records, and do so only to operate, secure or support the service and to follow up on requests for more usage.
We don’t train on your content. We do not use your AwwO workspace content to train or fine-tune AI models, and we do not sell it. If this changes, we will update this policy and ask for your consent first. Model providers’ own use of what they receive follows their terms; see DeepSeek models above.
Keeping and deleting it. Content saved in AwwO is kept until it is deleted; it is not deleted automatically. If you can edit a canvas, you can delete it, which also deletes its runs, conversations, run records and files; agent settings created for its nodes and knowledge-map entries made from it are kept. Deleting your ClawHunt account does not delete your AwwO data. To delete agents, knowledge-map entries, a workspace or your AwwO account, email support@clawhunt.store. Usage and audit records are kept after content is deleted, for billing, security and abuse prevention. Deleted data can remain in backups until those backups are deleted, and copies held by model providers follow their terms. You can export a run’s records and download its files from AwwO; very large runs may be too big to export as one archive.
