AI Signal 142
OtoDock ships self-hosted agent platform running Claude Code and Codex in departmental roles
OtoDock is a self-hosted, multi-tenant platform that deploys AI agents built on Claude Code and Codex, organized into company departments with configurable sharing modes and autonomous scheduling.
Only one feed is carrying this, so corroboration is limited. For teams already paying for Anthropic or OpenAI subscriptions, OtoDock offers a way to run agents on existing API keys without a separate SaaS bill, though the operational burden of self-hosting and sandboxing falls on the team.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Agents run on Claude Code, Codex, or local models using your own subscriptions and API keys, with per-agent and per-chat engine switching.
Four sharing modes control whether agents use private, shared, or hybrid workspaces across team members.
Agents can be scheduled, triggered by webhooks, given phone numbers via Twilio or Asterisk, and paired with remote machines through a single outbound connection.
THE READ
What the cluster adds up to.
OtoDock positions itself as a self-hosted company OS where AI agents act as digital employees organized into departments. Each agent is composed of six configurable parts, persona, memory, workspace, knowledge, skills, and tools, and can be assigned a role like Personal Assistant, System Admin, or Marketing Manager. The platform is multi-tenant by design, meaning multiple people interact with the same agents, and four sharing modes determine whether workspaces are private, shared, or both.
The platform runs on your own infrastructure, with agents executing in an isolated sandbox controlled by a single dashboard. You can pair additional machines, laptops, workstations, or PCs running macOS, Linux, or Windows, via a one-line install that maintains a single outbound connection to the server, requiring no open ports. If a paired machine goes offline, the server takes over. This architecture means teams retain control over where agent work happens and where files live, but it also means teams are responsible for maintaining the server and sandbox.
Agents are not limited to interactive chat sessions. They can run on schedules, fire on webhook events, or execute once at a chosen time, with every run producing a full chat log that can be reviewed and continued later. Notifications come in four severity levels. Agents can also be given phone numbers through Twilio, Asterisk, or FreePBX, allowing them to place and answer calls. The platform supports agent meetings where specialist agents discuss a topic together and converge on an answer in the dashboard.
The cost model is tied to existing subscriptions rather than a separate platform fee. Claude Code runs on your Anthropic subscription, Codex runs on your ChatGPT subscription, and local models on your own hardware are also supported. Every user connects their own subscription, and you can switch engines per agent or per chat. The project is described as fair source, though the specific license terms are not detailed in the available material.
Where this stops working is not explicitly bounded in the material, but the design implies several constraints. The platform depends on your ability to self-host and maintain a server with sandboxed agents. Phone integration requires a separate Twilio account or an existing Asterisk or FreePBX setup. The endless tools claim and document editing capabilities are listed as features but not demonstrated in detail in the available material. With only one feed carrying this story, there is no independent corroboration of the platform's claims or maturity.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER