# rivet-dev/sandbox-agent

Run Coding Agents in Sandboxes. Control Them Over HTTP. Supports Claude Code, Codex, OpenCode, and Amp.

Repository: https://github.com/rivet-dev/sandbox-agent
Canonical: https://ross.abutalabs.com/products/sandbox-agent
Homepage: https://sandboxagent.dev
Language: TypeScript
License: Apache-2.0
License Family: permissive
Topics: agent, ai, amp, claude, claude-code, codex, daytona, e2b, sandbox, opencode
Last push: 2026-06-19T06:05:41+00:00

## Health v2 (maintenance only)
Score: 69/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 88, release rhythm 76, longevity 15
- inputs: {"age_days": 220, "days_push": 75, "days_rel": 161, "gap_med": 1, "n_releases_24m": 20}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1549, forks 124 (observed 2026-08-28T04:05:01.924686+00:00)

## What it is
A server (with a TypeScript SDK) that runs inside a sandbox and exposes a universal HTTP/SSE API for controlling coding agents like Claude Code, Codex, OpenCode, Cursor, Amp, and Pi. It normalizes agent event formats into a universal session schema so events can be streamed, persisted, and replayed.

## Use cases
- run claude code inside an e2b sandbox and control it from my app over http
- swap between codex and opencode without rewriting my agent integration
- stream coding agent events to postgres and replay sessions later
- build a web ui that manages ai coding agent sessions remotely
- handle permission prompts from a coding agent running in docker
- persist agent transcripts for auditing and debugging

## When to choose
- you need to run coding agents in isolated sandboxes and control them remotely
- you want one unified API across multiple coding agents
- you need durable, replayable agent session transcripts

## When to avoid
- you only run a coding agent locally and interact with it directly
- you need a fully stable, production-hardened product (project is young)
- you use a coding agent not among the supported adapters

## Facets
- artifact type: library
- maturity: active
- function: agent-framework, http-server, sdk, api-framework
- domain: large-language-models, developer-tools, self-hosted
- platform: self-hosted, cross-platform, cli
- tags: coding-agents, sandbox, http-sse, claude-code, codex, opencode, session-replay, universal-adapter, ai-agents, docker, nodejs

## Member repositories
- rivet-dev/sandbox-agent (main) score 69

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:01.924686+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:30:24.119731+00:00, confidence not recorded.
  - readme: https://github.com/rivet-dev/sandbox-agent (fetched 2026-08-28T04:05:01.924686+00:00, sha 5e0a97612e8d)
  - homepage: https://sandboxagent.dev (fetched 2026-08-29T11:31:24.026948+00:00, sha 96ae3103dbf9)
  - site_page: https://sandboxagent.dev/docs (fetched 2026-08-29T11:31:24.036273+00:00, sha c0710c029c79)
  - registry_npm: https://registry.npmjs.org/sandbox-agent (fetched 2026-08-29T11:31:24.038013+00:00, sha 562b7dfc6d01)
- Data as of 2026-08-30T08:39:29.467469+00:00.