v0.722  ·  macOS · Windows · Linux  ·  MIT License

A genome browser with executable AI analysis

CodeXomics combines interactive genome visualization with conversational agents that execute real tools against your loaded data, from BLAST and primer design to protein structure lookup.

↓ Download Read the Docs → View Source
CodeXomics ChatBox
Find all DNA polymerase genes in this genome.
search_features → navigate_to_position → get_track_data
Found 12 DNA polymerase genes — jumped to dnaE at 3,842,101 bp.
Design primers around lacZ and check them on the sequence track.
design_primers → render on sequence view
Designed 3 primer pairs (Tm 58–62 °C), rendered on the track.
1desktop workspace
AIthat executes tools
3platforms supported
MCPclient integration
MITopen-source license
Capabilities

A full genomics workbench,
driven by conversation

One desktop workspace combining a genome browser, a multi-agent AI assistant, a dynamic tool registry, and a plugin marketplace.

🧬

Interactive Genome Visualization

High-performance SVG and Canvas rendering of genes, sequences, reads, variants, GC tracks, and custom annotations — zoom from chromosome to base pair.

💬

AI ChatBox

Multi-provider LLM configuration with dynamic tool injection. The assistant doesn't just answer — it executes tools against your live genome session.

🤖

Multi-Agent Runtime

Seven specialized agents coordinate analysis, data, navigation, external APIs, plugins, and deep research — routed automatically per task.

🧰

Dynamic Tool Registry

A structured registry retrieves relevant tools for each query, keeping prompts focused while the assistant works against the active genome session.

🔌

MCP Server

Model Context Protocol server with tools mode and agent mode over HTTP/SSE (port 3002) and WebSocket (port 3003) — drive CodeXomics from any MCP client.

🧩

Plugin Marketplace

VS Code-inspired extension architecture with activation events, contribution and command registries, marketplace support, and security validation.

Use Cases

From file to finding

Start with the question you have. CodeXomics keeps the data, view, and analysis connected as you work.

Explore a genomeLoad FASTA or GenBank, search features, and move from chromosome scale to base resolution.
Inspect evidenceAdd reads, variants, WIG tracks, or annotations and compare them in one synchronized view.
Run an analysisAsk for BLAST, GC content, primer design, protein structures, or literature-backed research.
Review and shareSave a project, export results, extend with plugins, or connect an MCP client.
Architecture

Seven agents, one coordinator

Each request is routed to the agent built for it — from quick navigation to autonomous deep research.

CoordinatorAgent

Orchestrates multi-step workflows and routes tasks to the right specialist.

AnalysisAgent

Sequence analysis, BLAST, primer design, GC content, and statistics.

DataAgent

File loading, track management, and data transformation.

NavigationAgent

Genome navigation, zoom, search, and view state control.

ExternalAgent

PubMed, preprints, AlphaFold, and external database lookups.

PluginAgent

Plugin discovery, installation, and marketplace operations.

DeepResearchAgent

Citation-verified deep gene research with archived, auditable proposals.

In Action

See it working

Real views from the app — every one of them reachable by asking the assistant.

Genome browser overview
Genome browser: feature track, GC content and skew, multi-track WIG data, gene details sidebar, and the protein/sequence view.
Aligned reads and references
Base-resolution and coverage read alignment tracks alongside gene references.
AlphaFold protein structure viewer
AlphaFold lookup and interactive 3D protein viewer, driven from the ChatBox.
Read alignment pileup
Base-level sequencing-read pileup over a selected gene.
Primer design on the sequence track
Designed primers rendered directly on the sequence view.
Compatibility

Speaks your file formats

FASTA GenBank GFF / GTF BED VCF SAM / BAM WIG KGML .prj.GAI projects
Get Started

Up and running in minutes

  1. Download the latest build from GitHub Releases, or build from source with Node.js 20/22.
  2. Configure an AI provider in Options → Configure LLMs — OpenAI, Anthropic, Google, or a local model.
  3. Load a genome via File → Load File or open a .prj.GAI project.
  4. Ask the assistant to navigate, visualize, and analyze — it executes real tools against your session.
  5. Extend it with plugins from the marketplace, or drive it from any MCP client.
# build from source
$ git clone https://github.com/Scilence2022/CodeXomics.git
$ cd CodeXomics && npm install
$ npm start
 
# standalone MCP server
$ npm run mcp-server