An agent sees only the tools the preset lists and you leave on. A tool that is off is not listed at all, so it costs the agent no context, and no agent can turn one on
Toolsets
| Toolset | What its tools do |
|---|---|
| Core | Start a task, find files and every copy of a path, prove types, read the docs, prepare an edit and check it. Always on |
| Code intelligence | Types, symbols, modules, the dw API, cartridges and site paths |
| Storefront | Routes, hooks, forms, resources, ISML, metadata, step types, pipelets, pipelines, content links, site archives and Page Designer |
| SCAPI and OCAPI | Custom APIs, and SCAPI and OCAPI types and contracts |
| Instance | Connections, code versions, WebDAV reads and compares, and raw instance logs |
| Log Center | Check for errors, parse and tail logs, find anomalies, write reports and mark issues fixed |
| Debugger | Drive your script and pipeline debug session, see Debugging with Agents |
| Ship | Diff, upload, deploy, code versions, WebDAV writes and the tool that runs plans |
| Remote data | Jobs, site imports and exports, clearing the page cache and On-Demand Sandboxes |
| Admin | Cartridge links, the active connection, code version and site scope, deploy settings, overrides and skills |
Risk
Every tool has a risk. The risk decides which presets list the tool
| Risk | What it covers |
|---|---|
| Project read | Reads the project in the IDE |
| Instance read | Reads an instance over the network |
| Local write | Changes in the project or on this computer that you can undo, such as a report or an issue's state |
| Instance write | Changes on an instance that another upload can reverse, and moving a live debug thread |
| Live change | Deletes, activations and other live changes |
| Runs code | Runs code on the instance |
Plan tools only prepare a change, so they read like Plans: Live change. The plan runs with intellijsfcc_mutation_execute, see Plans & Safety
Presets
Pick the preset in the list at the top of the Tools tab. It sets two things: the toolsets agents see, and the highest risk they may reach
| Preset | Toolsets | Maximum risk | What agents can do |
|---|---|---|---|
| Router core | Core | Read only | The smallest set, for agents with tight tool limits |
| SFCC read only | Core, Code intelligence, Storefront, SCAPI and OCAPI, Instance, Log Center | Read only | Read the project and your instances, and parse and tail logs. The preset B2C MCP starts with |
| SFCC operational | The same, plus Debugger | Local writes | Also set issue states, write reports, download files and read the debug session |
| SFCC with plans | All | Plans | Prepare any change and show you the plan. Nothing runs |
| SFCC full | All | Everything | Also run the plans they prepare |
| Custom | Custom toolsets: in Settings | Custom maximum risk: in Settings | Your own mix. Changing any tool of a preset switches to it |
The SFCC presets also list every read tool, whatever its toolset, such as intellijsfcc_deploy_status, intellijsfcc_job_status and intellijsfcc_ods_list. Router core and Custom list only their own toolsets
These rules hold under every preset:
- A tool that is off, or not in the preset, is not listed. A call to it fails as an unknown tool
- No agent can widen the toolsets or the risk. When a task needs more, the agent tells you which toolset, and you decide
- A connected agent's tool list changes only when you change the preset or the tools that are on
- An agent's own tools and instances in Connect Agent can only narrow this, see Connect Agents
The Tools tab
At the top: the preset list and Block Writes Now, then Search tools, Show and Expand All and Collapse All. Tools lists every tool under its toolset
- A toolset row reads 12/16 for the tools on, with a checkbox for the whole toolset. The checkbox shows a dash when only some are on
- A tool row shows its short name, without
intellijsfcc_, its risk, its calls in the last 7 days and its checkbox. A warning icon marks a tool that needs attention: Fails often, Refused often, Often cut, Answers over budget or Slow. Hover a row for its title and why - Click a checkbox, or select rows and press
Space, to turn tools or whole toolsets on or off. A checkbox clicked in a selection turns every selected row. The change reaches connected agents at once - Show narrows the list: All Tools, On, Off, Writes Only or Needs Attention. Search tools matches names, titles and descriptions
- A tool the preset leaves out reads Not in the preset. Some tools are not in the preset. Turning one on switches to a Custom preset
- While Block Writes Now is on, write tools read Writes blocked and their checkboxes are dimmed
- Right-click a tool for Turn On or Turn Off, Copy Tool Name and Copy Description
Select a tool to see its details, exactly as agents see them:
| Detail | What it shows |
|---|---|
| State: | Whether agents see it |
| Risk: | Its risk, from Project read to Runs code |
| Annotations: | The hints agents read: read only, destructive, idempotent, open world |
| Answer budget: | The most text one answer may hold |
| Last 7 days:, Latency:, Answer size: | Its calls, timings and sizes |
| Attention: | Why it needs attention, if it does |
| Input Schema | The arguments it takes |
Select a toolset to see how many of its tools are on. On a wide tool window the details sit beside the list, on a narrow one under it
| Toolbar button | What it does |
|---|---|
| Reset to Preset | Turns every tool of the preset back on |
| Export Tool List | Copies the tools that are on as a Markdown table |
Tools you turn off are saved in Tools turned off: in Settings, for every project
Block Writes Now
Block Writes Now on the Tools tab leaves agents only the read tools of your preset, at once, until you turn it off or restart the IDE. Queued writes don't start, and plans that haven't run can't run
Tools then reads Block Writes Now is on, so only read tools are available. To cut every agent off instead, click Stop