Skip to content

VS Code Setup ​

VS Code with GitHub Copilot supports native custom agents via .github/agents/ — .agent.md files with YAML frontmatter, each running in its own isolated context. VS Code also reads .claude/agents/ for Claude-format compatibility.

.claude/agents/ also works

VS Code detects .md files in .claude/agents/ following the Claude sub-agents format. If you already set up KAIROS for Claude Code, VS Code picks up the same files automatically.

Priority for deduplication: .github/agents/ > .claude/agents/

Prerequisites ​

Step 1 — Copy agents to .github/agents/ ​

VS Code uses .agent.md extension, but also detects plain .md files in .github/agents/:

bash
mkdir -p .github/agents
copy agents\*.md .github\agents\

Or use .claude/agents/ if you already have Claude Code set up.

Project layout with native VS Code format:

your-project/
├── .github/
│   └── agents/
│       ├── orchestrator-agent.agent.md
│       ├── pm-agent.agent.md
│       ├── architect-agent.agent.md
│       ├── context-extractor-agent.agent.md   <- Pre-pipeline: full-repo context (standalone)
│       ├── implementer-tdd-agent.agent.md    <- TDD implementer (default)
│       ├── implementer-coder-agent.agent.md  <- Code-first implementer (code, then the tests the project calls for)
│       ├── code-reviewer-agent.agent.md
│       ├── security-reviewer-agent.agent.md  <- Adversarial security review (optional, read-only)
│       ├── test-verifier-agent.agent.md
│       ├── qa-plan-agent.agent.md
│       ├── release-planner-agent.agent.md
│       ├── documentation-agent.agent.md      <- Feature-facing docs (optional, Phase 6b)
│       ├── retrospective-agent.agent.md      <- Standalone, post-pipeline: lessons capture
│       ├── improvement-advisor-agent.agent.md <- Standalone, infrequent: framework change proposals
│       ├── impact-assessment-agent.agent.md  <- Pre-pipeline: issue grounding, facts for the derivation (dispatched by the orchestrator)
│       ├── bug-triage-agent.agent.md         <- Bug reproduction + root cause (standalone)
│       └── dependency-audit-agent.agent.md   <- Standalone, periodic: dependency + tech-debt backlog

Step 2 — Agent file format ​

VS Code native .agent.md format supports additional fields beyond Claude's format:

yaml
---
name: PM Agent
description: Collects and structures requirements. Use at the START of a new feature.
tools: ['read', 'edit', 'search']
model: sonnet
handoffs:
  - label: "✅ Continue to Architecture"
    agent: architect-agent
    prompt: "Design the architecture for the requirements above."
    send: false
  - label: "⛔ Stop here"
    agent: ""
---

The handoffs field implements the HITL pattern natively: after the PM Agent responds, VS Code shows buttons ✅ Continue to Architecture and ⛔ Stop here. Selecting one switches context to the next agent with pre-filled prompt — no manual copy-paste needed.

VS Code maps Claude tool names automatically:

Claude toolsVS Code equivalent
read_fileread
write_fileedit
bashexecute (shell)

Step 3 — Configure handoffs for HITL ​

Add handoffs to each KAIROS agent file to create the HITL gate chain:

pm-agent.agent.md excerpt:

yaml
handoffs:
  - label: "✅ Approve → Architecture"
    agent: architect-agent
    prompt: "Design the architecture for these requirements: {output}"
    send: false
  - label: "✏️ Request changes"
    agent: pm-agent
    prompt: "Revise the requirements based on this feedback: "
    send: false

Repeat for each agent in the chain, pointing to the next phase.

Step 4 — Start a KAIROS session ​

Open Copilot Chat, select the Orchestrator agent from the dropdown (or type @orchestrator), and prompt:

Help me implement [your feature] using the KAIROS framework

The orchestrator delegates to PM Agent. After PM Agent responds, the handoff buttons appear — select ✅ Approve → Architecture to continue.

Step 5 — .kairos/ persistence ​

VS Code agents with edit in their tools list can write files. Add to each agent's instructions:

markdown
After the user approves, save the output to `.kairos/<feature_folder>/01-requirements.md`.

Or save manually by copying the Markdown output from the chat after each approved phase.

Feature comparison vs Claude Code ​

FeatureClaude CodeVS Code + Copilot
Native subagent format✅ .claude/agents/✅ .github/agents/*.agent.md
Context isolation per agent✅✅
Auto-delegation via description✅✅
HITL built-in gate✅ Orchestrator-enforced✅ Via handoffs buttons
.kairos/ persistence✅ Automatic⚠️ Via agent instructions or manual
Claude format compatibility✅ Native✅ Reads .claude/agents/*.md

Troubleshooting ​

ProblemFix
Agents not in dropdownCheck .github/agents/ has .agent.md files with valid frontmatter
Handoff buttons not appearingVerify handoffs: is in agent frontmatter and agent tool is in tools:
Wrong agent formatPlain .md files work in .github/agents/ but .agent.md unlocks full features
.kairos/ not writtenAdd edit to agent tools: and explicit instruction to write the file

Suggested Models ​

VS Code uses model IDs from the Copilot model picker. The model: field in agent frontmatter sets the model for that specific agent regardless of what is active in the chat panel.

Same reasoning/execution split as the shipped agents/*.md frontmatter — see Customizing models for the full rationale.

AgentRecommendedNon-Claude equivalent
orchestrator-agentopuso1, o3, Gemini Ultra
architect-agentopuso1, o3, Gemini Ultra — system design needs strongest model
context-extractor-agentopuso1, o3, Gemini Ultra
impact-assessment-agentopuso1, o3, Gemini Ultra — its facts drive the derivation of every downstream agent
security-reviewer-agentopuso1, o3, Gemini Ultra — adversarial analysis needs strongest model
improvement-advisor-agentopuso1, o3, Gemini Ultra
bug-triage-agentopuso1, o3, Gemini Ultra
pm-agentsonnetGPT-4o, Gemini 1.5 Pro
implementer-tdd-agentsonnetGPT-4o, Gemini 1.5 Pro
implementer-coder-agentsonnetGPT-4o, Gemini 1.5 Pro — no test-first overhead
code-reviewer-agentsonnetGPT-4o, Gemini 1.5 Pro
test-verifier-agentsonnetGPT-4o, Gemini 1.5 Pro
qa-plan-agentsonnetGPT-4o, Gemini 1.5 Pro
release-planner-agentsonnetGPT-4o, Gemini 1.5 Pro
documentation-agentsonnetGPT-4o, Gemini 1.5 Pro
retrospective-agentsonnetGPT-4o, Gemini 1.5 Pro
dependency-audit-agentsonnetGPT-4o, Gemini 1.5 Pro

Layered Mode not supported

VS Code does not run KAIROS Layered Mode (agents/team/, called Team Mode until v9.5.0). The orchestrator offers it only on Claude Code, and these agents are not mirrored for VS Code. Use the single implementer.

KAIROS Framework v9.7.1 — Released under the AGPL-3.0 License.