Building blocks for agent systems.
Atomic, Unix-like services — each does one job, small and focused — that agents drive through APIs and CLIs. Grouped by what they do: compute, data, intelligence, coordination, and operations. Pick the ones you need and follow each product’s setup. Authentication, quotas, and retention differ by service.
Where agents run — sandboxes, hosted runtimes, and browsers.
Sandboxes
forkable sandbox for agents.
Create Linux Boxes from images or filesystem Snaps. Fork one Snap at a time; persistent Volumes are separate. Execute through the CLI or SSH, within project quotas.
run9 → smith9 · availableAgent runtime
agent as a service.
Hosted agent Envs and persistent Sessions on run9. Configure a provider, then submit Turns; an Env can inherit project provider settings or override them.
smith9 → web9 · plannedBrowser
managed browsers for agents.
Planned managed browser sessions for agent workloads. Product scope and release timing are not finalized.
web9 →State that outlives a session — databases, storage, memory, secrets.
Postgres database
serverless Postgres, from terminal.
Serverless Postgres with database branching and built-in file storage. Create and query databases from the CLI with credentials scoped to your workload.
db9 → drive9 · availableObject storage
any sandbox, one filesystem.
Persistent files across sandboxes and sessions, with FUSE mounts, semantic file search, and explicit vault grants. Configure the context used by each agent.
drive9 → mem9 · availableAgent memory
one memory, many agents.
Persistent memory across supported agents and sessions, with tools to inspect, import, and export it. Follow the runtime integration or API setup guide.
mem9 → vault9 · plannedSecrets
secrets, sealed and granted.
Planned standalone credential management. Current run9 managed egress and drive9 vault grants have separate security boundaries.
vault9 →The model layer — gateways, skills, and evaluation.
Model gateway
one endpoint for every model.
Planned model gateway for routing and provider configuration. Product scope and release timing are not finalized.
gate9 → hub9 · plannedSkills & tools
a registry for agent skills.
Planned registry for agent skills and tools. Product scope and release timing are not finalized.
hub9 → eval9 · plannedEvals
offline evals, online quality.
Planned evaluation tools for agent workloads. Product scope and release timing are not finalized.
eval9 →How agents work together — queues, realtime, teamwork, schedules.
Queues
atomic task claim.
Shared tasks with dependencies and atomic claims. Optional pulse9 owner-liveness releases offline owners during subsequent task reads or claims.
task9 → inbox9 · pulse9 · tape9 · availableRealtime
messaging, presence, streams.
inbox9 messages, pulse9 worker leases, and tape9 append-only logs. Configure each service and implement scheduling, acknowledgements, and recovery in your application.
inbox9 · pulse9 · tape9 → chord9 · availableTeamwork
local and cloud agent workbench.
A workbench for native coding-agent Sessions on local or cloud Hosts. Organize work in Workspaces and Projects, continue conversations, and share artifacts.
chord9 → cron9 · plannedScheduling
cron and triggers for agents.
Planned scheduling and event triggers for agent workloads. Product scope and release timing are not finalized.
cron9 →Run it in production — observability and identity.
Agent transcripts
native agent session transcripts.
Capture native sessions and linked sub-sessions after configuring a supported runtime integration and space. Enable collection for the sessions you choose.
owl9 → auth9 · plannedAuth
one identity, any sys9.
Planned identity service. Continue using current product authentication and application authorization.
auth9 →task9 for task claims, inbox9 for messages,
pulse9 for worker leases, and smith9 for hosted Sessions.
Your application connects these APIs and implements scheduling, acknowledgements, retries,
and recovery. This is an architecture pattern, not an automatic integration.
— the composition story Start with one service. Compose the rest.
Install the CLI and follow the setup guide for your first service. Agents use the same commands with the credentials and permissions you provide.