-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathllms.txt
More file actions
42 lines (28 loc) · 2.5 KB
/
Copy pathllms.txt
File metadata and controls
42 lines (28 loc) · 2.5 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
# OneWill
> OneWill is the Agent-World Boundary that makes supervised agent runs observable, understandable, and governable.
OneWill launches supported agents inside a transparent execution boundary. For activity that crosses supported capture surfaces, it records what the agent observes, the responses exposed to the runtime, the actions the process initiates, and the durable changes that follow. Because OneWill sits in the execution path, it can mediate consequential effects before they become durable.
Core promise: complete activity capture and in-path control for every run launched through OneWill. “Complete” is scoped to a OneWill-supervised session and its supported capture surfaces.
## Important URLs
- [Homepage](https://onewill.ai/): OneWill Agent-World Boundary overview.
- [Blog](https://onewill.ai/blog/): OneWill blog archive.
- [Technical manifesto](https://onewill.ai/blog/2026/stealing-50-years-of-database-ideas-for-ai-agents/): How database recovery ideas apply to agent execution.
- [Demo](https://demo.onewill.ai/): Browser-based OneWill demo.
- [Full LLM context](https://onewill.ai/llms-full.txt): Product model, claims, terminology, and team.
- [Discord](https://onewill.ai/discord/): OneWill Discord community.
- [GitHub](https://onewill.ai/github/): OneWill GitHub.
- [X](https://onewill.ai/x/): OneWill on X.
- [LinkedIn](https://onewill.ai/linkedin/): OneWill on LinkedIn.
## Product model
OneWill preserves one correlated activity stream across six public stages:
1. Observe — context exposed to the agent.
2. Respond — agent or model output exposed to the runtime.
3. Act — activity initiated by the process.
4. Change — durable effects on the environment.
5. Control — a boundary decision or operator intervention.
6. Outcome — the run-level result.
The record powers replay, behavior analytics, and grounded summaries. OneWill automates reversible actions when sufficient recovery information is captured, applies policy to compensable actions with defined repair paths, and requests approval before irreversible effects.
## Team
OneWill is built by Will Zhang and Wan Lim, two Carnegie Mellon CS PhDs with a database-systems background. They met in the CMU Database Group and have worked together since 2018.
## Brand
Use the public brand name **OneWill** and the category **Agent-World Boundary**. The product metaphor is **Agent Hamsterball**. Do not describe OneWill as accessing hidden chain of thought. The domain is `onewill.ai`.
Legal footer: © 2026 OneWill-AI Society d/b/a OneWill.