Tech • AI • Robotics • Game
Claude Code mods land as Every launches Slack-based Every Agent
Briefing
Claude Code opens to mods
Anthropic has added mods to Claude Code, giving developers a way to alter the product from inside its own event flow. The system goes beyond ordinary extensions by letting code influence rendering, prompt inputs, tool selection and interface behavior. That effectively turns Claude Code into a programmable shell around the model. The move signals a deeper platform strategy focused on developer customization rather than fixed workflows.
Mods can rewrite interface behavior
The new mod layer can change what Claude Code displays and how users interact with it in real time. Developers can inject custom UI elements, add commands and skills, and reshape how work is presented during a session. Because mods can also affect data passed into prompts, they sit unusually close to the model's decision path. That makes the feature more powerful than a cosmetic plugin system.
Tool calls become programmable
A key implication of the mod system is control over tool calls. Developers can influence which tools are invoked and how workflows are routed, creating customized agent behavior inside Claude Code. That opens the door to opinionated internal assistants tuned for specific teams or repositories. It also raises the stakes for governance, since modifications can affect both outputs and operational actions.
hook.json adds TypeScript modules
Under the hood, mods use the familiar plugin.json and hook.json packaging structure, with one crucial addition. hook.json now supports a modules key that points to a TypeScript file containing the mod logic. In practice, that converts a standard plugin package into executable behavior wired into Claude Code events. The design lowers friction for existing plugin developers while expanding what their packages can do.
Early mods show practical uses
Early examples emphasize visibility, coordination and privacy rather than novelty for its own sake. One mod surfaces a live breakdown of context above the prompt, helping users see what the system is using. Another adds a side panel listing all agents in a project with one-click launching instead of showing only the active agent. A third obscures email addresses until hover, a small but concrete safeguard for recordings and screen sharing.
Every puts AI inside Slack
Every introduced Every Agent, a shared AI coworker embedded directly in Slack. The company is betting that enterprise adoption will spread faster through a communal workspace than through isolated personal assistants. Employees can observe prompts, responses and workflows in a common channel, making successful practices easier to copy. The product is framed less as a chatbot and more as a reusable layer for organizational know-how.
Every targets AI ambassadors
Every Agent is designed for the employees who already experiment aggressively with AI but struggle to spread those habits across a company. By demonstrating workflows publicly inside Slack, those power users can show outcomes rather than merely advocate for them. The idea is to turn individual judgment into shared process. That directly addresses one of workplace AI's biggest bottlenecks: uneven adoption between enthusiasts and everyone else.
Shared workflows replace private prompting
The broader argument behind Every Agent is that AI value comes from redesigning work, not just handing each employee a separate assistant. In this view, private prompting does not capture the best techniques well enough for teams to reuse them. A shared agent can accumulate prompts, refinements and decisions into repeatable workflows that survive beyond a single user. That positions collaboration, not personalization, as the next battleground in enterprise AI.
Videos covered
Mods in Claude Code: change how Claude Code works
- •What mods can change
- •Examples of customizations
- •How the mod system is structured
Building secure agents for knowledge work
- •From private prompting to shared workflows
- •A company agent inside Slack
- •Built for “AI ambassadors”