sandbox_exec
Run a REAL command in the user's active session sandbox (isolated cloud shell + live MCP). kind='bash' (command) | 'mcp' (endpoint, action list|call, tool, arguments) | 'agent' (instruction). Needs wallet_address; anon works. No active sandbox → says so.
This record as markdown: /tools/ai-rokha-rokha/sandbox-exec.md
What sandbox_exec does on Rokha
AI agents invoke sandbox_exec to trigger actions in Rokha. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call: builds kicked off, notifications sent, workflows started.
| Parameter | Type | Required | Description |
|---|---|---|---|
kind | string | Yes | |
tool | string | — | |
image | string | — | |
action | string | — | |
params | object | — | |
command | string | — | |
endpoint | string | — | |
arguments | object | — | |
instruction | string | — | |
wallet_address | string | Yes |
Parameters from the server's own tool schema.
Why sandbox_exec is rated High
Executes arbitrary bash commands, MCP actions, and agent instructions with real-world effects in user sandbox.
From the tool's definition Run a REAL command in the user's active session sandbox; bash, mcp, agent
Risk signalsAccepts URL/endpoint input (endpoint) · High parameter count (10 properties)
Attacks that exploit this kind of access
The rule that runs sandbox_exec safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Rokha, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For sandbox_exec, this is the rule to start with:
sandbox_exec stays usable, but rate-capped: a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Rokha, apply this rule, and every sandbox_exec call is checked against it from then on.
Questions about sandbox_exec
Run a REAL command in the user's active session sandbox (isolated cloud shell + live MCP). kind='bash' (command) | 'mcp' (endpoint, action list|call, tool, arguments) | 'agent' (instruction). Needs wallet_address; anon works. No active sandbox → says so. It is categorised as a Execute tool in the Rokha MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
sandbox_exec accepts 10 parameters: kind, tool, image, action, params, command, endpoint, arguments, instruction, wallet_address. Required: kind, wallet_address. The full parameter table on this page comes from the server's own tool schema.
Register the Rokha MCP server in PolicyLayer and add a rule for sandbox_exec: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Rokha. Nothing to install.
sandbox_exec is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the sandbox_exec rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for sandbox_exec. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
sandbox_exec is provided by the Rokha MCP server (https://rokha.ai/mcp/jsonrpc). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Rokha, and thousands of servers like it.
Across the catalogue