Agent tools
Overview
The in-pod agent has access to 55+ tools for interacting with the pod, your code, and the Meshia platform. Tools are grouped by category.
Filesystem
| Tool | Description |
|---|---|
| read_file | Read a file's contents |
| write_file | Create or overwrite a file |
| edit_file | Make targeted edits to an existing file |
| list_directory | List files and folders in a directory |
| search_files | Search file contents with regex |
| glob_files | Find files matching a glob pattern |
| delete_file | Delete a file (with confirmation) |
| move_file | Move or rename a file |
| file_info | Get file size, permissions, modification time |
Shell and execution
| Tool | Description |
|---|---|
| bash | Run a shell command and return output |
| python | Execute a Python snippet |
| background_task | Run a long command in the background |
| kill_process | Stop a running background process |
| list_processes | Show running and background processes |
Git
| Tool | Description |
|---|---|
| git_status | Show working tree status |
| git_diff | Show file diffs |
| git_commit | Stage and commit changes |
| git_branch | Create or switch branches |
| git_push | Push commits to remote |
| git_pull | Pull latest from remote |
| git_log | Show commit history |
| create_pr | Open a GitHub pull request |
Experiments
| Tool | Description |
|---|---|
| log_metric | Log a metric value at a step |
| log_artifact | Save a file as an experiment artifact |
| create_experiment | Start a new tracked experiment |
| end_experiment | Mark an experiment as complete |
| list_experiments | List experiments in the session |
| compare_experiments | Generate a comparison of two or more runs |
Data
| Tool | Description |
|---|---|
| download_url | Download a file from a URL |
| download_hf_dataset | Download a Hugging Face dataset |
| download_hf_model | Download a Hugging Face model |
| upload_artifact | Upload a file to Meshia storage |
Environment
| Tool | Description |
|---|---|
| pip_install | Install Python packages |
| conda_install | Install packages via conda |
| gpu_status | Get GPU utilization, VRAM, temperature |
| system_info | Get CPU, RAM, disk, OS details |
| env_var | Set an environment variable |
UI and canvas
| Tool | Description |
|---|---|
| canvas_add | Add an element to the canvas |
| canvas_update | Update an existing canvas element |
| canvas_remove | Remove a canvas element |
| render_chart | Generate a chart from data |
| render_pie_chart | Generate a pie or donut breakdown from category data |
| render_gpu_partition | Show GPU pod slots grouped by workload |
| render_table | Generate a sortable table |
| render_diff | Show a side-by-side diff |
| show_image | Display an image from the pod |
Research
| Tool | Description |
|---|---|
| search_papers | Search Semantic Scholar for papers |
| read_paper | Download and extract text from a paper |
| create_plan | Build a research plan |
| update_plan | Modify an existing plan step |
| memory_store | Save a fact to agent memory |
| memory_recall | Retrieve relevant memories |
Communication
| Tool | Description |
|---|---|
| notify_user | Send a notification (email or in-app) |
| post_blackboard | Post a message to the session blackboard |
| read_blackboard | Read recent blackboard posts |
This isn't exhaustive, the agent has additional internal tools for coordination and planning. But these are the ones you'll see referenced in logs and can ask the agent about directly.