1mcp FAQs

Question 1

How does 1mcp achieve such high token reduction?

Accepted Answer

Instead of the LLM making multiple individual tool calls, 1mcp allows agents to write JavaScript/TypeScript code that chains MCP calls together. This entire chain is executed in a single, sandboxed environment, drastically cutting down on the tokens required to orchestrate complex operations.

Question 2

Can 1mcp be integrated with existing AI SDKs?

Accepted Answer

Yes, 1mcp features seamless integration with popular AI SDKs via the `@1mcp/ai-sdk` package. This allows you to easily convert your existing AI SDK tools into sandboxed MCP tools, maintaining a consistent execution story across your backend and browser.

Question 3

What differentiates 1mcp from other cloud-based sandboxing solutions?

Accepted Answer

1mcp offers superior token efficiency through chained execution (matching Anthropic's 96% reduction), uses standard JS/TS with npm dependencies (no vendor lock-in), provides dynamic session-scoped security policies, and supports browser-side execution to offload server compute, unlike global settings found in other solutions.

Question 4

What security measures does 1mcp provide for code execution?

Accepted Answer

1mcp executes code in a secure, sandboxed WASM/QuickJS runtime. It offers granular security policies for each session, including strict network access controls (allowed/denied domains), filesystem permissions (read-only/writable paths), and runtime limits (memory, timeout, stdout).

Question 5

What is 1mcp and its primary benefit for AI agents?

Accepted Answer

1mcp empowers AI agents to compose and execute sandboxed code for chained Model Context Protocol (MCP) tool calls. Its primary benefit is significantly reducing token usage (up to 96%) by avoiding context bloating common in sequential tool calls.

1mcp

About

Key Features

Use Cases