Centroid definition

The Default VS Code AI Coding-Agent Extension

Denominator: In-cluster, N=22. Source: Manual-audit/CENTROID-VS-CODE-AI-CODING-AGENT-EXTENSIONS.md. Reader meaning: this page explains the sector default, not a recommendation that every project should copy it.

Definition

The synthetic centroid is the composite default project assembled from recurring traits in the accepted corpus. For this sector, the default project is a TypeScript or JavaScript VS Code extension that presents an AI coding assistant through a sidebar, chat, or webview surface.

It is package-backed for Marketplace/Open VSX-style installation, fresh and non-archived, individual- or small-team-led, and framed around coding-agent behavior: file/workspace context, edits, tools, local or external runtime bridges, and configurable model providers.

MCP and ACP are not universal centroid traits. Local protocol/client bridges are common enough to shape the sector edge, so they appear through the broader C8 local tool/process bridge trait.

Sector Boundary

Inclusion Rules

  • Public GitHub repository with source-visible VS Code or VS Code-compatible extension package evidence.
  • The headline product is an AI coding assistant, coding agent, coding-agent client, or IDE-native coding workflow surface.
  • Repository is non-archived, fresh within the SC-021 180-day window, and package-backed.
  • AI coding functionality is the product surface, not only a generic sample, prompt list, telemetry layer, or unrelated extension.

Exclusion Rules

  • Closed-source commercial-only agents.
  • Editor forks whose product is the whole IDE rather than the VS Code extension surface.
  • Generic VS Code extension samples or starters without AI coding-agent behavior.
  • Standalone CLIs, apps, or web services with no VS Code package surface.
  • MCP servers, ACP runtimes, observability tools, skills managers, and memory layers unless they also ship the coding-agent or coding-agent client extension surface.

Centroid Traits

Denominator: In-cluster, N=22. Source: centroid artifact trait table. Reader meaning: a trait qualifies when it appears in more than half of the accepted corpus.

TraitNamePrevalencePercentDescription
C1VS Code package surface22/22100%Repo contains a VS Code or VS Code-compatible package manifest with engines.vscode and contribution metadata.
C2TypeScript / JavaScript extension implementation19/2286%The extension implementation is primarily TS/JS/Node rather than a non-extension wrapper.
C3Sidebar, chat, or webview IDE surface17/2277%Primary user surface is an IDE sidebar, chat UI, webview, or equivalent extension UI.
C4AI coding assistant or coding-agent headline22/22100%README or manifest positions the project around coding assistance, code-agent behavior, or AI coding workflow support.
C5Individual or small-team maintenance16/2273%Ownership is individual-led or small-team rather than purely official platform SDK.
C6Multi-model or BYO provider posture15/2268%Project supports multiple model/provider options, local providers, or OpenAI-compatible/BYO configuration.
C7Agentic file/workspace operation16/2273%Project claims or implements file edits, command execution, codebase context, debugging, tool use, or task workflow beyond static chat.
C8Local tool/process bridge14/2264%Project bridges to MCP, ACP, CLI runtimes, local models, terminal/debugger/tool processes, or external coding-agent runtimes.
C9Recent active source22/22100%Repository is non-archived and pushed within the SC-021 180-day freshness window; most are much fresher.
C10Marketplace/installable extension package22/22100%Project ships as a VS Code/Open VSX/VS Code-compatible extension package rather than only docs or a standalone service.

Trait Application by Repo

Denominator: In-cluster, N=22. Source: centroid artifact trait-application table. Reader meaning: matched-trait counts explain CM scoring inputs without changing final calibration scores.

RepoMatched traitsCountEvidence summary
cline/clineC1, C2, C3, C4, C5, C6, C7, C8, C9, C1010/10Open-source coding agent in the IDE and terminal, with file edits, command execution, browser use, MCP, SDK, CLI, and VS Code package surfaces.
continuedev/continueC1, C2, C3, C4, C5, C6, C7, C8, C9, C1010/10Open-source AI code agent with VS Code extension, CLI, autocomplete, rules, context providers, model brokering, and source-controlled checks.
Kilo-Org/kilocodeC1, C2, C3, C4, C5, C6, C7, C8, C9, C1010/10AI coding-agent platform with VS Code package, autocomplete, tool use, gateway/SDK packages, and broad agent UI surface.
zgsm-ai/costrictC1, C2, C3, C4, C5, C6, C7, C9, C109/10Enterprise-oriented strict AI coder extension with VS Code app/package, webview UI, core packages, evaluation packages, telemetry, and CLI-adjacent surfaces.
codota/tabnine-vscodeC1, C2, C4, C5, C7, C9, C107/10Extension-specific Tabnine repo for AI chat, autocomplete, documentation, tests, and enterprise privacy/compliance positioning.
DeepMyst/MystiC1, C2, C3, C4, C5, C6, C8, C108/10Multi-agent VS Code coding team that coordinates Claude Code, Codex, Gemini, and Copilot-style agents.
10Nates/ollama-autocoderC1, C2, C4, C5, C8, C106/10Local Ollama autocompletion extension with options and streaming behavior.
Lianues/Lim-CodeC1, C2, C3, C4, C5, C6, C8, C108/10VS Code AI programming assistant with multi-channel model support, tool calling, and modular architecture.
andrepimenta/claude-code-chatC1, C2, C3, C4, C5, C6, C8, C108/10Claude Code chat interface for VS Code plus an MCP permissions package.
adifyr/shadow-codeC1, C2, C4, C5, C6, C106/10VS Code/Cursor extension centered on pseudocode-to-production-code workflows.
rusiaaman/chat.mdC1, C2, C3, C4, C5, C8, C107/10Markdown-file-first LLM chat interface with MCP/tool integration and editable transcripts.
vscode-reborn-ai/vscode-reborn-aiC1, C2, C3, C4, C5, C6, C8, C108/10BYO-AI VS Code coding assistant with local LLM support and general code/refactor/improve workflows.
cogflows/promptcode-vscodeC1, C2, C4, C5, C85/10Prompt/code rescue tool available as VS Code extension and CLI for when code agents hit limits.
ZhouChaunge/DeepCopilotC1, C2, C3, C4, C5, C66/10Low-star DeepSeek-powered coding agent embedded in VS Code with file editing, terminal, search, and planning claims.
Kile-Thomson/Rokket-GSDC1, C2, C3, C4, C5, C6, C8, C108/10VS Code frontend for the GSD AI coding agent with chat UI, multi-model support, visualizations, and workflow claims.
jasonjmcghee/claude-debugs-for-youC1, C2, C4, C5, C85/10VS Code extension that lets an MCP client debug code with breakpoints in-editor.
Hive-Academy/ptah-extensionC1, C2, C3, C4, C5, C6, C8, C108/10Provider-agnostic AI orchestration for VS Code with workspace intelligence, multi-agent workflows, and MCP server integration.
AizenvoltPrime/damoclesC1, C2, C3, C4, C5, C6, C8, C108/10Claude AI coding assistant using an Agent SDK with subagent/tool visualization and a Python voice sidecar.
justimyhxu/claude-code-local-for-vscodeC1, C3, C4, C5, C8, C106/10Claude Code VS Code extension patch adding local and remote mode selection.
formulahendry/vscode-acpC1, C2, C4, C5, C85/10Agent Client Protocol client for VS Code that connects to ACP-compatible agents.
joaompfp/hermes-vscodeC1, C2, C3, C4, C5, C86/10VS Code sidebar for the Hermes AI agent over ACP, with streaming chat, tools, sessions, and context tracking.
Harsh1210/openclaude-vscodeC1, C2, C3, C4, C5, C6, C8, C108/10OpenClaude VS Code assistant powered by any LLM, with sessions, MCP IDE server, permissions, diff, worktree, and webview components.

Validation

All 10 traits clear the greater-than-50% threshold against the N=22 corpus. The lowest-prevalence centroid trait is C8 local tool/process bridge at 14/22, which remains above threshold. The strongest traits are C1, C4, C9, and C10 at 22/22 because they come directly from the validated corpus boundary.

Sensitivity noteReader meaning
ACP-only clientsIf excluded, C8 would remain above threshold through MCP, CLI, local-model, and tool bridges, but the sector would lose an important protocol-client edge.
Lineage-bearing Claude/Cline derivativesIf excluded, the centroid would not materially change, but the role mix would shift away from patch-level long-tail client surfaces.
Metadata-only Issue #19 candidatesIf added, stale archived projects would distort C9 and adoption metrics; SC-024 keeps Issue #19 as scout context only.

Source Artifacts

Source map for this page. Denominator and trait text come from the centroid artifact; role and score effects are cross-checked against calibration and visual data.

Published artifactSource roleVisible metric families
Manual-audit/CENTROID-VS-CODE-AI-CODING-AGENT-EXTENSIONS.mdCanonical centroid definitionSector definition, inclusion rules, trait prevalence, trait application, validation notes.
Reports/VISUAL-DATA-VS-CODE-AI-CODING-AGENT-EXTENSIONS.mdChart-ready trait tableCentroid trait prevalence counts and percentages.
Manual-audit/CALIBRATION-TABLE-RADAR-VS-CODE-AI-CODING-AGENT-EXTENSIONS.mdScore and role cross-checkCM scores and public role effects derived from centroid evidence.
Operations/Sprint-Contracts/SC-024.mdReport-generation scopeAccepted corpus and source-refresh context.
Operations/SECTOR-RADAR-DISPLAY-STANDARD.mdDisplay contractDenominator, Source, Reader meaning, and correction-route requirements.

Corrections

Found a stale repo, wrong source trace, or boundary classification issue? Send a correction to hello@diversum.dev with the repo name, displayed value, and source that should replace it.