the universal sandbox runtime for agents and humans.
pocketenv.io
sandbox
openclaw
agent
claude-code
vercel-sandbox
deno-sandbox
cloudflare-sandbox
atproto
sprites
daytona
Refactor README Key Features section
+23
-37
+23
-37
README.md
+23
-37
README.md
···
24
24
25
25
## Key Features
26
26
27
-
### Secure isolated environments
28
-
29
-
Run commands, agents, and code safely in isolated sandboxes.
30
-
31
-
### Vendor-agnostic abstraction
32
-
33
-
Pocketenv can run sandboxes on different providers transparently:
34
-
- Daytona
35
-
- Cloudflare Sandbox
36
-
- Vercel Sandbox
37
-
- Deno Sandbox
38
-
- And more to come!
39
-
40
-
No provider lock-in.
41
-
42
-
### Agent-native design
27
+
- **Secure Isolated Environments**
28
+
Run commands, agents, and code safely inside isolated sandboxes with strong separation.
43
29
44
-
Pocketenv is designed for AI workflows:
45
-
- Run OpenClaw
46
-
- Run Codex CLI
47
-
- Run Claude Code
48
-
- Run Gemini CLI
49
-
- Run Copilot CLI
50
-
- Run custom agents
30
+
- **Vendor-Agnostic Abstraction**
31
+
Run sandboxes transparently on Daytona, Cloudflare Sandbox, Vercel Sandbox, Deno Sandbox — and more coming.
32
+
No provider lock-in.
51
33
52
-
### Open sandbox definitions via AT Protocol
34
+
- **Agent-Native Design**
35
+
Built for modern AI workflows:
36
+
- OpenClaw
37
+
- Codex CLI
38
+
- Claude Code
39
+
- Gemini CLI
40
+
- Copilot CLI
41
+
- …and your custom agents
53
42
54
-
Sandboxes are defined using open lexicons and stored on [AT Protocol](https://atproto.com) PDS.
43
+
- **Open Sandbox Definitions via AT Protocol**
44
+
Sandboxes defined with open lexicons and stored on AT Protocol PDS. Enables discoverable, shareable, versioned sandboxes and a growing community ecosystem.
55
45
56
-
This enables:
57
-
- Discoverable sandboxes
58
-
- Shareable sandboxes
59
-
- Versioned sandboxes
60
-
- Community ecosystem
46
+
## Example Use Cases
61
47
62
-
### Example Use Cases
63
-
- Prototype AI agents safely
64
-
- Run Codex / Claude / Gemini in isolation
65
-
- Test untrusted code securely
66
-
- Share reproducible dev environments
67
-
- Run temporary dev environments
68
-
- Provide sandbox infrastructure for your app
48
+
- Prototype and run **OpenClaw** agents safely in isolated environments
49
+
- Prototype AI agents in complete safety
50
+
- Run Codex / Claude / Gemini / OpenClaw code in isolation
51
+
- Securely test untrusted or third-party code
52
+
- Share reproducible developer environments
53
+
- Spin up temporary / ephemeral dev sandboxes
54
+
- Provide sandbox infrastructure as a service for your application