Agent Tools | Omnilib Docs

Omnilib's agent tools let the AI assistant perform actions on your behalf inside your project. When you use Agent mode in AI Chat, the AI can read and write files, run commands, search content, manage your knowledge graph, compile LaTeX, and much more.

Understanding Agent mode

When you use Agent mode:

Describe what you want to accomplish.
The AI decides which tools to use and shows you its plan.
You approve or reject each action before it executes.
The AI executes approved actions and continues.

This keeps you in control while letting the AI handle multi-step tasks.

Approval workflow

For each tool call, you see:

Tool name — which action the AI wants to take
Parameters — the inputs it will use
Preview — what will happen

Click Approve to execute or Reject to skip. After a rejection, the AI can suggest an alternative approach.

Some actions — like overwriting files or running shell commands — show an extra confirmation prompt.

File tools

`file.read`

Reads the contents of a file in your project. Can read specific line ranges for large files.

`file.write`

Creates a new file or overwrites an existing one. Shows a preview of the content before writing.

`file.edit`

Makes a targeted change to an existing file — finds a specific string and replaces it. Shows a before/after diff before executing.

Code execution

`bash.execute`

Runs a shell command and returns stdout and stderr to the AI. Useful for running tests, build scripts, installing dependencies, or any other command-line task.

Commands execute in your project directory. A configurable timeout (default 30 seconds) prevents runaway processes.

Search

`vector.search`

Searches your project files using semantic vector similarity. Finds content related to a natural language query even when the exact words do not appear in the files. Returns ranked results with file paths and relevant snippets.

LaTeX

`latex.compile`

Compiles a LaTeX document using Tectonic. If compilation produces errors, the AI iteratively fixes them and recompiles until the document builds successfully or the error cannot be resolved automatically.

Knowledge graph

The graph tools let the AI read and navigate your project's knowledge graph — the network of files, concepts, and relationships Omnilib builds from your content.

Tool	What it does
`graph.summary`	Returns a high-level summary of the knowledge graph structure
`graph.get_structure`	Returns the full graph schema: node types and edge types
`graph.find_related`	Finds nodes related to a given node by following edges
`graph.search`	Searches the graph by label or property value
`graph.get_node`	Returns detailed information about a specific node

Bibliography

The bibliography tools manage references and citations in your research documents.

Tool	What it does
`bibliography.search`	Searches your reference library by title, author, or keyword
`bibliography.add`	Adds a new reference to the project bibliography
`bibliography.cite`	Inserts a formatted citation for a reference into the active document

Project memory

The memory tools let the AI read and update the project memory — the structured context it uses to give consistent, project-aware responses across conversations.

Tool	What it does
`memory.query`	Asks a question answered from project memory
`memory.add`	Adds a new fact or note to project memory
`memory.entities`	Lists all named entities tracked in memory
`memory.stats`	Returns statistics about memory size and coverage
`memory.entity_facts`	Returns all known facts about a specific entity

Jupyter notebooks

The notebook tools let the AI read and execute cells in Jupyter notebooks open in your project.

Tool	What it does
`notebook.read_cells`	Returns the content and output of notebook cells
`notebook.add_cell`	Inserts a new code or Markdown cell at a specified position
`notebook.modify_cell`	Edits the source of an existing cell
`notebook.start`	Starts the notebook kernel
`notebook.stop`	Stops the notebook kernel
`notebook.execute_cell`	Runs a specific cell and returns its output

File tagging

Tool	What it does
`tags.get`	Returns all tags applied to a file
`tags.update`	Sets or replaces the tags on a file
`tags.search`	Finds files that have a specific tag

Task tracking

Tool	What it does
`task.ask`	Pauses the current task and asks you a clarifying question before continuing
`task.track`	Records a task or sub-task for the AI to track across multiple steps

Comments

`comments.manage`

Lists, creates, resolves, and replies to comments in your project. The AI can use this tool to add review comments to files, mark comments as resolved, or reply to existing threads.

Security

All agent tools are scoped to your project files and run under your user permissions. The AI cannot access files outside your project directory, and every action requires explicit approval before execution. All tool calls are logged for audit purposes.