AI assistance, inside Spyder

Spyder AI Chat Plugin

Work with cloud or local language models without leaving your IDE. Stream rich answers, share project context, run guarded agentic tools, and complete code directly in the editor.

Install from PyPI Project page
  • Spyder 6.1+
  • Python 3.9+
  • MIT licensed
  • No added runtime dependencies
12built-in providers
Native + Texttool transports
Local + Cloudmodel freedom
Project-awareSpyder integrations

A real IDE workspace

Stay in the flow from question to verified change.

The dockable chat pane lives beside the editor, console, Variable Explorer, Plots, and Code Analysis. Responses render incrementally, with code, thinking, and tool activity kept readable as they arrive.

  • Ask with contextAttach code, plots, runtime summaries, or project files.
  • Review in placeKeep explanations and proposed actions beside the source.
  • Verify immediatelyUse the Console and analysis panes without leaving Spyder.

See it working

A complete workflow inside Spyder.

Watch the plugin move between conversation, project context, generated changes, and Spyder's native panes in a real session.

Watch directly on YouTube

One pane, broad capability

Built for practical Python work.

Use the model as a conversational partner, a context-aware reviewer, or a guarded project agent.

01

Streaming conversation

Incremental Markdown, code blocks, tables, nested lists, collapsible thinking, image input, saved prompts, commands, and searchable chat history.

02

Project context

Attach selected code, whole files, project folders, unsaved editor buffers, console output, images, and authorized external directories.

03

Agentic tools

Let capable models inspect and modify files, run Python and Git, lint code, make HTTP requests, and return results for the next reasoning step.

04

Editor completion

Use FIM ghost-text completion with local or cloud endpoints, configurable triggering, provider-specific templates, and familiar accept or dismiss controls.

05

Runtime insight

Preview bounded, type-aware Variable Explorer summaries and deterministic before-and-after baselines without transferring complete objects by default.

06

Reusable skills

Expose personal or project Agent Skills progressively, with separate current-chat and future-chat availability plus project-scoped trust.

Agentic, with boundaries

Inspect. Plan. Change. Verify.

Native function calling and the portable Text DSL share the same confirmation, execution, history, and security path. You decide which tools exist and whether modifying actions require approval.

  1. Model proposesTool calls appear as readable, typed actions.
  2. You stay in controlReview batches or allow configured autonomous execution.
  3. Results returnOutputs feed back into the chat for the next step.

Spyder-native context

Connected to the tools already around your code.

Editor

Send a selection or file, insert generated code, and reload files changed by agentic actions.

IPython Console

Share console output and run reviewed Python actions in the active project session.

Variable Explorer

Redact and preview capped metadata, samples, statistics, and baseline comparisons.

Plots

Attach plot images and context for explanation, comparison, or improvement.

Projects

Bind context and modifying tools to the open Spyder project root.

Code Analysis

Run Pylint through the agent and surface reports in Spyder's analysis pane.

Configuration without guesswork

Choose the provider. Define the boundaries.

Provider, reasoning, tool transport, permissions, confirmations, and diagnostics stay visible and configurable inside Spyder.

Provider controlConfigure hosted or local endpoints, credentials, models, and thinking behavior.
Explicit permissionsSelect execution mode, confirmation policy, and every tool exposed to the model.

Choose the model, not the lock-in

Twelve providers out of the box.

Use hosted APIs, your own inference server, or any compatible custom endpoint.

Bring your own model access. The plugin does not include free hosted AI usage. Commercial providers require your own API key and may charge for requests; alternatively, connect to a model running locally on your computer through Ollama, LM Studio, vLLM, or another OpenAI-compatible server.

  • OpenAI
  • Groq
  • Mistral
  • DeepSeek
  • Together AI
  • Fireworks AI
  • OpenRouter
  • Azure OpenAI
  • Ollama
  • LM Studio
  • vLLM
  • Custom

OpenAI-compatible Chat CompletionsStreaming, vision where supported, reasoning fields, and native tool calls.

Portable Text DSLAgentic workflows for compatible models without native function calling.

Local-first friendlyAPI keys are optional for endpoints that do not require authentication.

Power needs clear edges

Tools stay inside an explicit security model.

Project-root protectionModifying actions require an open, trusted Spyder project and cannot silently escape its root.

Configurable approvalManual, Semi, and Full modes let you choose when action batches pause for confirmation.

Visible communicationDiagnostic traces can record sanitized request and execution lifecycles without adding FIM traffic.

Local model supportKeep inference on your machine with Ollama, LM Studio, vLLM, or a custom compatible server.

Ready in a few commands

Install into Spyder's Python environment.

Activate the environment that runs Spyder, install the package, then restart the IDE. The pane is available under Window > Panes > AI Chat.

PyPI package SourceForge project Full documentation
Terminal
python -m pip install spyder-ai-chat

Spyder AI Chat Plugin logo

Keep the assistant close to the work.

Open source, provider-flexible, and designed specifically for Spyder.

Get Spyder AI Chat
Expanded Spyder AI Chat screenshot