What Is Consoles?
An AI workspace for every person in your organization. Connected to the tools they already use, with the control, visibility, and governance enterprises actually require.
The market for AI assistants has a coverage problem.
ChatGPT and Claude.ai work well for individuals. Cursor and GitHub Copilot work well for engineers. Neither was designed for the operations manager tracking vendor contracts, the finance analyst pulling from three systems, or the support lead summarizing tickets. And none of them were designed for the IT team that has to decide whether to let any of this run on production data in the first place.
Consoles is built for both groups.
— partial support ✕ not available
| Capability | Consumer AI | Dev tools | Consoles |
|---|---|---|---|
| Individual use | |||
| Non-developer users | |||
| Connect any tool | |||
| Org-level skills | |||
| Audit trail | |||
| SSO + SCIM | |||
| Per-user isolation | |||
| VPC / on-prem deploy | |||
| Multi-model support |
What it is
Consoles is a desktop AI workspace. Every person in a company gets their own agent environment, connected to the tools they already use: Salesforce, Slack, Linear, Jira, Google Drive, internal APIs, private databases. They research, analyze, automate, and complete work through conversation.
It runs on Mac, Windows, and Linux. It supports every major AI model: Claude, GPT-4o, Gemini, and open-source alternatives. Users switch models mid-session.
The interface works like email and a task manager. No terminal. No JSON. No code unless you want it.
The foundation
Consoles is built on Craft Agent, an open-source framework that wraps Anthropic’s Claude SDK and the Pi agent SDK. This is the same architectural relationship Cursor has with VS Code: a solid foundation, extended with a product layer aimed at a specific market.
Craft Agent handles the session loop, tool routing, and agent runtime. What we built on top: multi-tenant org management, enterprise identity (SSO, SCIM, RBAC), gated skill and source publishing, per-user VM isolation, audit logging across every tool call, and the deployment options (VPC, on-prem) that enterprise procurement actually requires. The agent is the foundation. The org layer is the product.
Connecting tools
Most AI tools make you configure integrations manually. OAuth credentials, JSON schemas, API documentation. Fine for a developer. A barrier for everyone else.
In Consoles, you say “connect to Salesforce” and the agent handles the OAuth flow, reads the API documentation, tests the connection, and generates the context it needs to use the tool effectively. This works for MCP servers, REST APIs, GraphQL endpoints, local filesystems, and private internal services.
Once a source is set up, admins can publish it org-wide. A sales rep opens Consoles and their CRM is already there, preconfigured, scoped to their team. They didn’t configure anything.
Control and observability
This is the part that matters to the org, not just the individual user.
Every tool call Consoles makes is logged: who made the request, which source it hit, what was returned, when. Admins have a full audit trail across the entire organization. When a compliance review asks what your AI systems touched, the answer is already there.
Admins control which connections are available to which teams, what data flows through them, and what the agent is allowed to do. Connections are scoped to specific groups. Access is revoked the moment someone is offboarded through your identity provider.
Skills, the reusable instruction sets that define how the agent behaves, are published and governed centrally. Legal publishes a contract review workflow. Finance publishes an expense analysis template. The org controls what agents do, not just whether they can run.
Credentials are encrypted at rest (AES-256) and in transit (TLS 1.3) and never enter conversation context. Tokens live in the OS keychain and are injected at the API boundary. The agent cannot see them.
For organizations that cannot accept data leaving their infrastructure: Business plan users get a dedicated VM per seat. Enterprise deploys into your own VPC or on-prem. Traffic stays inside your perimeter.
SSO via Okta, Azure AD, and Google Workspace. SCIM provisioning syncs users and groups from your directory. RBAC. Role-scoped visibility. Onboarding and offboarding happen once, in one place.
Who it’s for
The analyst, PM, ops manager
Works across tools without needing a developer to set anything up.
- Say "connect to Salesforce" — it connects
- Switch AI models mid-session
- Org-published skills available automatically
- No terminal, no config files, no code
- Interface works like email and a task manager
IT, security, leadership
Full control over what AI agents can access and do across the org.
- Every tool call logged: who, what, when
- Admin-gated skills, sources, and connections
- RBAC, per-team scoping, group policies
- SSO (Okta, Azure AD, Google) + SCIM
- VPC or on-prem deploy — data stays in your infra
These two groups have different needs. Consoles is designed to satisfy both at once. The employee gets a capable agent. The organization gets the controls required to let that agent run.
Pricing
Solo use is free, no credit card required. Team plans start at $25 per user per month and add org-level skills, shared sources, SSO, and multiplayer sessions. Business is $39 per user per month with dedicated infrastructure per seat, SCIM, RBAC, and audit logs. Enterprise pricing is custom and includes VPC or on-prem deployment.
Consoles is built on the Craft Agent open-source framework.