Documentation

Build a tenant-safe agent in 15 minutes.

Install the CLI, spawn an agent scoped to one tenant with a hard cost cap, and watch cross-tenant access fail at the broker. Full reference docs are rolling out per product — start here.

terminal

# 1 · install
$ curl -fsSL https://sys9.ai/install | sh

# 2 · spawn a tenant-scoped agent, capped at $5
$ sys9 agents spawn --tenant acme \
    --runtime claude-code --cap $5.00
  → agent ag_7Qx2 ready in 480ms
  → identity oidc:acme/ag_7Qx2 · budget $5.00

# 3 · cross-tenant access is impossible
$ sys9 agents spawn --tenant acme --as-tenant globex
  → broker: 42501 tenant_mismatch

Start by product

SandboxIsolated execution for untrusted agent code. RuntimesClaude Code, Codex, OpenClaw, Cursor — managed. DatabasePer-tenant Postgres with vector search. MemoryDurable agent memory: working, recall, archival. IdentityPer-agent OIDC and OAuth-on-behalf-of. GatewayThe broker that enforces every tenant boundary. DurableDurable execution for long agent workflows. ReplayTime-travel debugger for any agent run.

For AI agents

Point your coding agent at the skill file and it will install, authenticate, and use sys9 autonomously:

Read https://sys9.ai/skill.md and follow the instructions

Need something that isn't here yet? hello@sys9.ai or request access.