Understand the mental model
Think of 100xprompt as a capable engineer at a well-stocked workbench. Every action goes through a tool - one well-defined capability. Two things bound each tool:- Permissions - what you’ve allowed. Sensitive actions like editing files or running commands can prompt for approval, run freely, or be blocked outright.
- The agent’s allowed set - each agent is granted only the tools its job needs. A read-only research agent has no editing tools on its bench.
You never call tools yourself. Describe what you want in plain language and 100xprompt picks the right tools to get there - asking for permission where your settings require it.
See the toolbox at a glance
Know what each tool does
Here’s the full toolbox in plain terms - each capability and what it lets 100xprompt do for you.Work with files
| Capability | What it lets 100xprompt do |
|---|---|
| Read a file | Open any file in your project - code, config, docs, or images - to understand what’s there before acting. |
| Write a file | Create a brand-new file, or fully replace an existing one, with the exact contents you asked for. |
| Edit in place | Make surgical, targeted changes to an existing file without rewriting the whole thing - a single replacement or every match at once. |
| Edit notebooks | Add, change, or remove cells in a Jupyter notebook while preserving its structure. |
Find things
| Capability | What it lets 100xprompt do |
|---|---|
| Find files by name | Locate files by name or pattern - the answer to “where does X live?” |
| Search text in code | Scan your codebase for a string, symbol, or pattern and return every match with context. |
| List a directory | Browse a folder as a tree to get oriented in an unfamiliar part of the project. |
| Understand code semantically | Resolve a symbol’s actual definition, references, and implementations - plus its type info and call hierarchy - using language intelligence, not just text matching. |
| Look up library docs | Pull up-to-date documentation and code examples for third-party libraries, SDKs, and public APIs - the reference material your local codebase doesn’t contain. |
Run commands
| Capability | What it lets 100xprompt do |
|---|---|
| Run a command | Execute shell commands - build, test, lint, run scripts, use git, or drive your own CLI tooling - and read back the output. |
Reach the web
| Capability | What it lets 100xprompt do |
|---|---|
| Fetch a page | Retrieve a specific URL and read its contents - docs, an API reference, a changelog. |
| Search the web | Look up current information across the web and pull back relevant results. |
Delegate work
| Capability | What it lets 100xprompt do |
|---|---|
| Launch a subagent | Hand a focused sub-task to a specialized agent - for example, a broad read-only search - so the main thread stays clean. |
| Check on a subagent | Follow a running subagent’s progress and collect its results when it finishes. |
Stay organized and check in
| Capability | What it lets 100xprompt do |
|---|---|
| Track a task list | Maintain a live checklist for multi-step work so you can watch progress and nothing gets dropped. |
| Ask you a question | Pause and ask for a decision or clarification when a task has a genuine fork in the road. |
Extend the toolbox
| Capability | What it lets 100xprompt do |
|---|---|
| Invoke a skill | Load a Skill - a packaged set of instructions and helpers - to handle a specialized job well. |
| Schedule a task | Set work to run automatically on a schedule you define. |
| Remember & forget | Save durable facts and preferences to long-term memory, and retract them when they change. |
| Connect external tools (MCP) | Bring in tools from external systems - issue trackers, design tools, browsers, databases - through MCP. |
See how a tool is gated
Two independent gates decide whether a tool is available at a given moment. A tool call passes the agent’s allowed set first, then your permissions.Gate 1 - the agent's allowed set
Every agent carries its own bench of tools. A research-focused agent may be read-only, with no way to write files or run commands. When you delegate work, the subagent inherits only the tools its role needs.
Work well with tools
Let 100xprompt search before it edits
Let 100xprompt search before it edits
You rarely need to point at exact files. Describe the change and 100xprompt finds the right place with its search tools, then edits precisely.
Approve commands thoughtfully
Approve commands thoughtfully
The command tool is the most capable and the most consequential. Use Permissions to pre-approve safe, routine commands and keep a prompt on anything that changes state.
Delegate big searches
Delegate big searches
For sprawling “look everywhere” questions, 100xprompt spins up a subagent that sweeps many files and reports back just the conclusion - keeping your main conversation focused.
Grow the toolbox as you need it
Grow the toolbox as you need it
Skills and MCP connections add new capabilities on demand. Start with the built-ins; reach for extensions when a task calls for them.
Related
- Connect tools with MCP: bring trackers, browsers, databases, and design tools into the toolbox.
- Skills: package instructions and helpers into reusable capabilities 100xprompt can invoke.
- Permissions: control what 100xprompt may run, edit, and reach.
Connect Tools with MCP
Bring external systems - trackers, browsers, databases, design tools - into the toolbox.
Skills
Package instructions and helpers into reusable capabilities 100xprompt can invoke.
Permissions
Control what 100xprompt is allowed to run, edit, and reach - with prompts, allowlists, and denials.