Extend what agents can do.

Five built-in extensions handle documents, code, terminals, application observation, and local inference. Each runs in isolation with its own tools and UI. Third parties build on the same SDK.

Code Intelligence

Index the codebase into a searchable structure.

Navigate dependencies, find references, analyse impact of changes. Agents query the index instead of reading raw files, reducing token usage by orders of magnitude.

Find all references, get definition, search by meaning
Interactive graph explorer with dependency visualisation
Semantic vector search with local embeddings
Cross-repo federation for multi-project dependencies

Structural indexingGraph navigationSemantic searchCross-repo20 languages

Documents

Read, write, and convert PDF, Word, Excel, and PowerPoint files locally.

Local processing for most formats, with optional deep parsing for scanned documents and complex tables. Agents handle documents as structured data, not raw text.

Tier 1: Fast local extraction for standard documents
Tier 2: Deep parsing for complex layouts, scanned pages, OCR
Tier 3: Remote processing for specialised formats (opt-in)
Drop files into any chat for automatic routing to the right processor

PDF, DOCX, XLSX, PPTX3 processing tiersLocal firstFile drop routing

Terminal

Detect available commands from the project configuration.

Manage running processes. Agents can execute operations with user approval, streaming results in real time.

Auto-detects scripts from project configuration files
Streaming output with progress indicators and results
Process lifecycle management: start, monitor, stop
Interactive shell sessions for complex operations

Command detectionProcess managementStreaming outputInteractive shells

Witness

Let agents observe running applications through three tiers.

Structured DOM extraction (most efficient), accessibility tree reading (native apps), and screenshot analysis (fallback). Evidence capture with timestamps for audit trails.

Tier 1: DOM extraction from web applications, ~50x cheaper than screenshots
Tier 2: OS accessibility tree for native apps on all platforms
Tier 3: Screenshot + vision model as fallback when structure is insufficient
Proof recording: each step captured with timestamp and tier used

DOM extractionAccessibility treeVision fallback~50x fewer tokens

Goose

Run AI models locally via Ollama with zero data leaving the machine.

Multi-provider fallback across 9+ vendors. Per-agent provider selection. Full autonomy mode for trusted operations.

Local inference with open-weight models, no API keys required
9+ providers: Anthropic, OpenAI, Google, Groq, Mistral, and more
Full-auto mode for autonomous execution without approval prompts
Per-slot configuration: different provider per specialist agent

Ollama (local)9+ providersFull-auto modePer-slot config

Build your own

Build extensions for your organisation.

The plugin SDK exposes 8 extension points: workspace pages, tab pages, configuration panels, sidebar entries, MCP tools, terminal tools, file handlers, and status displays. Plugins run in isolated child processes. Organisation policy controls which plugins are available to the team.

Five plugin types: CLI agents, MCP servers, daemons, exporters, and utilities. Each runs in full isolation with its own config namespace and lifecycle. Submit to the marketplace for third-party distribution.

Workspace page

Full-tab component with sidebar entry

Tab page

Dynamic plugin-hosted tabs in the layout

MCP tools

Tools callable by any agent in the workspace

Terminal tools

Shell commands agents can invoke with streaming

Config panel

Settings forms, setup wizards, status pages

File handler

Route file drops to the right processor

Agent vendor

Add a new LLM provider to the spawn dropdown

Status display

Icon and panel in the agent chat header

Local tools. No lock-in. No data leaks.

Every extension ships with the Studio. Build custom extensions with the same SDK.

Book a demo Back to Studio

Book a Call

See how Swarmix works

A walkthrough of unified visibility, governance, and intelligence across every AI vendor and team.

Deploy in weeks
Works with any existing AI stack
SOC 2 compliant, GDPR ready