01 available workspace
Work in OpenOrange
Run agents, collaborate in chat, inspect requests, and understand model and service usage from one team workspace.
$ workspace Private workspace for AI agents / staging active
OpenOrange Private AI workspace
Run agents, collaborate in Code, inspect every request, understand model and service usage, and keep access, privacy, and control in the same place.
OpenOrange product sequence
// one connected product
OpenOrange - Workspace
$ openorange open private workspace
Give each team one private place for users, agents, apps, chat, models, services, requests, and billing.
01 OpenOrange - Workspace
02 OpenOrange - Agents
03 OpenOrange - Code
04 OpenOrange - Usage
05 OpenOrange - Control
06 OpenOrange - Services
// product proof
One product, from daily work to the full picture.
The current workspace, active OpenOrange Code work, and incoming runtime integrations read as one coherent product while keeping current and incoming capabilities clear.
02 in review
Build in OpenOrange Code
Share coding projects and history while isolated sessions handle execution, forks, queues, handoff, models, and GitHub workflows.
$ openorange code 03 integrations underway
Bring the right agent
Use OpenClaw now, then add Hermes, IronClaw, and specialized runtimes as their integrations land.
$ runtime adapters surface state proof
workspace private by team Users, agents, apps, chat, models, services, billing, and requests share one clear instance boundary.
agent chat context + approvals Chat exposes runtime identity, streaming, queues, context usage, tool decisions, and session cost in the same conversation.
code projects + sessions OpenOrange Code is in review with shared projects, isolated sessions, durable queues, forks, writer handoff, model choice, and per-user attribution.
runtimes open, not locked OpenClaw runs today; Hermes and IronClaw are coming through explicit runtime integrations rather than product forks.
usage models + services Requests, tokens, cache, spend, search, mail, connectors, reliability, and limits remain attributable.
control policy + privacy Access grants, model limits, tool decisions, redaction state, contextual approvals, and audit history stay attached to the work.
OpenOrange system position
// system spec
Everything your team needs to run agents well.
OpenOrange stays simple: a private workspace, multiple runtimes, collaborative Code, connected services, clear usage, and controls that stay close to the work.
Private workspace
One dedicated OpenOrange instance for each team.
Users, agents, apps, flows, API keys, chat, Brain, models, services, billing, and requests live together without collapsing runtime data into a central SaaS account.
Agent runtimes
Choose the runtime that fits the work.
OpenClaw is integrated today. Hermes and IronClaw are on the way, with runtime cards and adapters keeping identity, chat, channels, health, and usage consistent across them.
OpenOrange Code
Collaborative coding agents inside the same private workspace.
Teams share projects, models, configurations, sessions, history, prompt queues, forks, and writer handoffs. Each session still runs in an isolated OCI container with its own toolchain.
Usage and observability
See what every person, agent, model, and service consumes.
Model spend, tokens, cache, requests, search, mail, connectors, latency, reliability, and limits stay attributable instead of disappearing into provider dashboards.
Access, privacy, and control
Controls appear where people need them, not as a separate product.
Manage user grants, model policy, limits, tool decisions, redaction, local secret references, contextual approvals, and attributable audit history from the private instance.
Apps and services
Useful capabilities around every agent runtime.
Connect apps, flows, browser access, search, mail, and other services while keeping provider state, activity, latency, reliability, and accounting visible.
// FAQ
FAQ
Frequently asked questions
OpenOrange is a private workspace where a team runs AI agents, collaborates in chat and Code, controls access and models, and understands requests, services, and spend.
A private instance brings users, agents, projects, chat, access, requests, usage, and billing views together. Agent runtimes keep their own execution state, memories, media, and local data.
Secrets, sessions, memories, media, caches, runtime databases, overlays, and raw logs stay on the instance or runtime host. OpenOrange works with scoped credentials, references, redacted observations, and attributable request metadata.
OpenClaw is integrated today. Hermes and IronClaw integrations are on the way, and the runtime boundary is explicit so specialized or custom agents can join the same workspace without pretending every runtime behaves identically.
Raw payloads do not need to become control-plane data. Request traces can preserve actor, route, model, token, cost, cache, redaction state, and audit context while sensitive contents stay local or redacted.
No. Ordinary work stays direct. Instance policy can allow, ask, or deny sensitive capabilities, so approval appears only when it adds value while attributable audit history remains available.
A private instance includes its own workspace, users, agents, chat, apps, flows, API keys, Brain, model access, services, request history, usage, billing views, privacy controls, and account settings.
Token values, provider keys, registry credentials, and channel credentials stay in local secret files or secret backends. OpenOrange stores secret references and audit context, not copied secret values.
The instance dashboard attributes model activity by actor, agent, route, model, request, tokens, cache behavior, price version, and tariff. Staging now extends the same operational view to service activity such as search, mail, connectors, and web access.
Request a private OpenOrange pilot
Intake brief received.
A human will review the brief and reply via your channel.