Marketplace

Ask questions, attach images, and let the assistant draft files and run multi-step tasks for you

See everything your extensions, nodes, agents, and MCP servers can do, in one searchable catalog

Edit code with syntax highlighting for every language you work in, plus breakpoint markers in the gutter

Hit Ctrl+Shift+P, start typing, and run any command without hunting through menus

Open CSV and TSV files as a real table, with the delimiter detected for you and a filter box on top

Understand a CSV or Parquet file at a glance: column stats, distributions, and how columns correlate

Connect to your PostgreSQL and MySQL databases, browse their schema, and run SQL against them

Adds MySQL to the Database extension so you can connect, browse schema, and run SQL

Adds PostgreSQL to the Database extension so you can connect, browse schema, and run SQL

Keep every dataset you work with in one list and pull new ones from Kaggle or Hugging Face

Browse and search Hugging Face datasets, then pull the ones you want into your workspace

Browse and search Kaggle datasets, then pull the ones you want into your workspace

Set breakpoints and step through your Python code, watching variables and the call stack as it runs

Manage your containers, images, volumes, and networks, and edit docker-compose files visually

Query your CSV and Parquet files with plain SQL, no database server to set up

Follow a course lesson by lesson, with a tutor that asks instead of answering

Read what a course teaches, module by module, before you commit to it

Find a course, see what it teaches, and download it into your workspace

Adds runnable lesson notebooks to the Learn extension, with cells your tutor can gate

Create and switch Python environments, install packages, and share the active one across extensions

Run eval suites over LLM-producing targets and gate on the scores

Scores LLM output as a manifest so eval suites can gate on structural validity

Scores a retrieval against the documents a case declared relevant so eval suites can gate on recall

Scores an LLM answer against the assertions its eval case declares so suites can gate on them

Package a governed run as a signed record of what ran, under which policy, and who approved it

Track your training runs on a local MLflow server and compare them side by side

Browse your workspace as a file tree, and open, create, rename, copy, cut, paste, or delete files and folders. Paste images and text straight from the system clipboard.

Powers the Generate commands by turning a goal into a finished IvyNode, IvyAgent or IvyMCP artifact

Runs attack packs against a prompt and answers whether the guardrail held

Generates red-team attacks and grades a guardrail verdict with promptfoo on this machine

View PNG, JPG, GIF, WebP, and SVG files with zoom, pan, and a checkerboard behind transparency

Edit the inputs for a node or agent run as JSON, starting from the example the node ships with

Open and run Jupyter notebooks with per-cell editing, streaming output, and a kernel picker

Build agent flows on a canvas, wiring nodes and MCP tools together, then run them

Package your nodes and tools into an MCP manifest that other MCP clients can call

Serve one of your MCP manifests as a real MCP server over stdio or HTTP

Build, run, and test your own reusable tools from code cells, with input schemas and test specs

Preview what is inside a node file, from its runtime and code cells to its tests, without editing it

Search published nodes, preview what they do, try them out, and rate the ones you use

Search, preview, and run every node you have created in this workspace

Check a kernel's details, its installed Python packages, and whether it is reachable

Connect Jupyter kernels and run your code on them from notebooks and flows

Start a local Jupyter server in your active Python environment with one click, CPU or GPU

Point at a remote Jupyter server, and have the connection checked before it is saved

Defines what a knowledge base contains and keeps its index current

View PCD and PLY point clouds in 3D with full color, and orbit, zoom, and pan around them

See where your data came from, with the stages between a dataset and the model it trained

Publishes this workspace's past lineage runs to any OpenLineage server that Marquez or DataHub can read

Describe your LLMOps process and have it written, then see every stage live on one map and run it

Define what your LLMs are allowed to do and manage those instruction classes in one place

Register the LLM models your team serves over the network so extensions can call them

Run GGUF models on your own machine through llama.cpp and send local LLM calls to them

Define, generate and test LLM instruction classes in an editor backed by workspace files

Find a model in Ollama's library or on HuggingFace, or point at a weights file on disk, and add it as a core

Install Ollama, download models into its store, and run them locally with nothing leaving the machine

Run ONNX models locally for chat, embeddings, and rerank through an onnxruntime server

Register the remote LLM services and endpoints you want to call, and keep them in one place

Inspect your instruction classes and try them out before wiring them into anything

Browse the extension registry and install, enable, disable, update, or remove what you find

Read .md, .markdown, and .mdx files rendered properly, with syntax-highlighted code blocks

See which tools your connected MCP servers offer, and run any of them from a generated form

Run MCP servers as local processes so their tools are callable everywhere you work

Watch MCP traffic as it happens, both the calls you make and the ones served to others

Connect to MCP servers over HTTP or SSE and make their tools available where you work

Connect and manage your MCP servers, local or remote, with the connections saved for next time

Build a training pipeline with gates, run it locally, and see exactly which gate stopped it

Adds DVC to the MLOps extension so you can version training data and read which data a run used

Adds drift checks to the MLOps extension, stopping the pipeline when your data has moved

Adds data contracts to the MLOps extension, stopping the pipeline when your data stops matching

Read your registered models and their champion through whichever registry engine this workspace has

Keep your models in one list, add them from Kaggle or Hugging Face, and run local ones

Browse and search Hugging Face GGUF models, then download the ones you want to run locally

Browse and search Kaggle models by task and modality, then pull them into your workspace

Watch your node, agent, and workflow runs as they happen, with logs and variables per cell

Ships run traces to an OpenTelemetry collector so your existing tracing tools can read them

Catch everything your extensions are telling you, with a status-bar bell for anything unread

Browse your buckets and move files in and out of S3, Azure Blob, Google Cloud, or MinIO

Adds Azure Blob Storage to the Object Storage extension so you can browse and move blobs

Adds AWS S3 to the Object Storage extension so you can browse buckets and move objects

Adds Google Cloud Storage to the Object Storage extension so you can browse and move objects

Adds MinIO and other S3-compatible stores to the Object Storage extension

Read the time-stamped log lines your extensions write, all in one panel

Export the workspace policy as something an external agent gateway can enforce

See every error and warning in one list, and click any of them to jump to the line

Edit requirements.txt and install, remove, or check each package against the environment you pick

Find and replace across one file or your whole workspace, with regex, case, and glob filters

Keep your API keys and tokens somewhere safe, in your OS keychain or in HashiCorp Vault

Adds an OS-encrypted local store to the Secrets extension, backed by your platform keychain

Adds HashiCorp Vault to the Secrets extension, storing your secrets as KV v2

Change any extension's settings from one place, and import or export the whole configuration

Stage, commit, and read your history as a branch graph in a Git sidebar next to your files

Run a hyperparameter sweep and see which point won, through the engine your workspace already has

Define the shell commands you run all the time and launch any of them from one panel

Run shell commands in a real terminal next to your work, with as many sessions as you need

Write test specs for your tools and run them with pass and fail results in one panel

Adds pytest to the Test extension so you can run your Python tests and gate a pipeline on them

A sidebar panel for appearance options, with the controls still on the way

Trace what the system is doing under the hood, check subsystem health, and export a redacted log

Turns your files into a searchable index that Knowledge and semantic search build on

Opens a detail page for any extension, with its stats, what it can do, and the actions you need

Open a folder to start work, or jump straight back into a workspace you had open