Claude Code ↔ Copilot CLI: Changelog-Driven Feature Parity

Changelog-driven feature parity is a methodology for mapping when each CLI tool first shipped a given capability and when its competitor matched it, using release notes rather than current documentation — revealing arrival order and gap duration that docs never show.

Why Changelog Tracking

Official documentation reflects current state, not arrival order. When you ask "does Copilot CLI support hooks now?" the answer from the docs is yes. What the docs don't show: Claude Code shipped hooks well before Copilot CLI matched with preToolUse in v0.0.396 (2026-01-27). That multi-month gap matters for your adoption planning.

Changelog-driven parity tracking surfaces:

• Which capabilities are now universal (available in both tools)
• Which remain tool-specific and in which direction
• The velocity at which the trailing tool closes a gap

Sources

• Claude Code changelog: code.claude.com/docs/en/changelog — covers v0.2.x through current
• Copilot CLI releases: github.com/github/copilot-cli/releases — machine-readable via gh api repos/github/copilot-cli/releases
• GitHub Changelog: github.blog/changelog/ — GA announcements and major feature posts

Methodology

This table tracks semantic equivalence rather than string matching. "Copilot Extensions" (v1.0.3) is not the same API as "Claude Code skills + hooks" — but they solve the same problem: bundling custom tools and lifecycle callbacks into an installable package. The table identifies functional equivalents and notes where implementations diverge.

Extraction approach:

1. Fetch both changelogs programmatically (gh api for Copilot CLI; WebFetch for Claude Code)
2. Categorize entries by capability domain (hooks, plan mode, memory, MCP, etc.)
3. Find the first version in each tool that ships each domain
4. Flag entries in the trailing tool that close a previously open gap

Feature Parity Matrix

As of 2026-03. Sources: Claude Code changelog; Copilot CLI releases v0.0.349–v1.0.3; GitHub Changelog.

Capability	Claude Code	Copilot CLI	Status
MCP server support	pre-2026 (exact version not in public changelog)	v0.0.337 (2025-10-08)	Both
Plan mode	pre-2026 (Shift+Tab)	v0.0.387 (2026-01-20, /plan)	Both
Skills / slash commands	pre-2026 (SKILL.md auto-discovery)	v0.0.371 (2025-12-18, ~/.copilot/skills/)	Both
Hooks (lifecycle interception)	pre-2026 (14+ documented events)	v0.0.396 (2026-01-27, preToolUse)	Both
Plugin / extension system	Skills + marketplace	v0.0.392 (2026-01-22, /plugin)	Both
Parallel sub-agents	pre-2026 (/batch)	v0.0.374 (2026-01-02, Explore/Task agents)	Both
Custom agent definitions	pre-2026 (.claude/agents/*.md)	v0.0.353 (2025-10-28, .github/agents/)	Both
Permission system / autopilot	pre-2026 (hooks block with exit code 2)	v0.0.351 (2025-10-24); v0.0.399 /allow-all	Both
Cross-session memory	v2.1.59 (/memory)	v0.0.384 (2026-01-16, store_memory tool)	Both
Context compaction	pre-2026 (auto-compact)	v0.0.374 (2026-01-02, 95% threshold)	Both
Multi-model support	v2.1.68 (model picker)	v0.0.341+ (Claude, GPT, Gemini)	Both
Remote / background delegation	pre-2026 (claude remote-control; /remote-control in v2.1.79)	v0.0.353 (/delegate); v0.0.384 (& prefix)	Both
Worktree isolation	pre-2026 (--worktree, WorktreeCreate/Remove hooks)	—	Claude Code only
Auto-decomposition into parallel worktrees	pre-2026 (/batch, 5–30 units)	—	Claude Code only
Agent teams (shared task list, direct messaging)	pre-2026, experimental	—	Claude Code only
HTTP hooks (POST/JSON via hooks)	v2.1.76+ (type: "http" on WorktreeCreate hook)	—	Claude Code only
Voice mode (STT, push-to-talk)	v2.1.69+ (20 languages)	—	Claude Code only
/loop with cron scheduling	pre-2026	—	Claude Code only
Extended thinking effort controls	v2.1.80+ (effort frontmatter field)	—	Claude Code only
Agent SDK (TypeScript + Python)	pre-2026	—	Claude Code only
/research command (exportable reports)	—	v0.0.417 (2026-02-25)	Copilot CLI only
/review command (built-in code review agent)	—	v0.0.388 (2026-01-20)	Copilot CLI only
/chronicle (standup / session history)	—	v0.0.419 (2026-02-27)	Copilot CLI only
MCP Elicitations (structured form input)	—	v0.0.421 (experimental)	Copilot CLI only
Extensions via @github/copilot-sdk	—	v1.0.3 (experimental, 2026-03-09)	Copilot CLI only
Session usage metrics (events.jsonl)	—	v0.0.422	Copilot CLI only
ACP server mode	—	v0.0.397	Copilot CLI only

Key Cross-Matches in Detail

Hooks

Claude Code ships a hooks system with 14+ lifecycle events (per the Claude Code changelog): PreToolUse, PostToolUse, Stop, SubagentStart, and others. An exit code of 2 from a hook blocks the tool call. This covers deny, modify, and notification patterns.

Copilot CLI v0.0.396 added preToolUse (deny/modify); v0.0.401 added agentStop and subagentStop; v0.0.422 added personal hooks and startup hooks. The Copilot CLI surface is approximately 5 confirmed events vs Claude Code's 14+ documented events (per the Claude Code changelog). The mechanisms are equivalent; the coverage differs.

Source: Copilot CLI releases; Claude Code changelog.

Skills / Extensions

Both tools ship a discoverable package format:

• Claude Code: SKILL.md files in skills/, auto-discovered; marketplace install via npx skills add owner/repo
• Copilot CLI v0.0.371: ~/.copilot/skills/ directory; /skills add in v0.0.396; plugin-provided skills in v0.0.394+; SDK-generated extensions in v1.0.3

The implementation differs: Claude Code uses declarative YAML/markdown; Copilot CLI v1.0.3 Extensions are code-generated via @github/copilot-sdk (experimental). Both bundle MCP servers, agent definitions, and lifecycle hooks.

Plan Mode

Both tools separate planning from execution:

• Claude Code: Shift+Tab enters Plan Mode, a read-only analysis mode before code changes
• Copilot CLI v0.0.387: /plan with a dedicated panel; v0.0.415 adds plan approval menu with autopilot and fleet options

Copilot CLI's plan mode (v0.0.387, January 2026) shipped after Claude Code's Shift+Tab plan mode, which predates the public changelog window (pre-2026), with a more structured approval UI.

Keeping the Table Current

Both changelog sources are machine-readable:

# Copilot CLI releases (JSON, paginated)
gh api repos/github/copilot-cli/releases --paginate

# Claude Code changelog (fetch and parse)
# https://code.claude.com/docs/en/changelog

Update cadence: Copilot CLI ships weekly (sometimes daily); Claude Code ships roughly weekly. Check both within a 2-week window to catch parity closures.

When you see a tool-specific feature appear in the other tool's changelog, update the Status column from "X only" to "Both" and add the version.

Key Takeaways

• As of 2026-03, the core agentic primitives — MCP, plan mode, hooks, skills, memory, parallel agents — are available in both Claude Code and Copilot CLI
• Claude Code leads on worktree isolation, /batch decomposition, extended thinking controls, and the Agent SDK
• Copilot CLI leads on built-in command agents (/research, /review, /chronicle), the ACP server mode, and SDK-generated extensions
• Copilot CLI closed most major gaps between late 2025 and early 2026 — a 4–6 month lead time from idea to parity across core primitives
• Semantic equivalence analysis matters more than keyword matching; "Extensions" and "skills + hooks" describe different implementations of the same capability

When This Backfires

Changelog-driven parity tracking is only as good as the changelogs themselves:

• Silent features: Both tools frequently ship capabilities without changelog entries or under generic labels like "minor improvements." Version-pinning a feature to a specific release may be incorrect if it landed earlier via a quiet update.
• Semantic equivalence disputes: Deciding that "Claude Code skills + hooks" is functionally equivalent to "Copilot CLI Extensions" requires judgment calls that other practitioners may contest. The table encodes one perspective; implementations differ in ways that matter for specific use cases.
• Staleness rate: Both tools ship weekly or faster. A parity matrix accurate today may misclassify features as "tool-specific" within a week. Treat the table as a snapshot requiring regular refresh, not a durable reference.

Example

Copilot CLI v0.0.422 (released 2026-02-14) includes the entry: "Add personal hooks and startup hooks."

Step 1 — Categorise. Domain: hooks. Capability: hooks (lifecycle interception).

Step 2 — Check the current matrix. The Hooks row shows Status = "Both", with Copilot CLI first matched in v0.0.396 (preToolUse). This release extends hook coverage; it does not open a new gap.

Step 3 — Update the detail section. In Key Cross-Matches in Detail → Hooks, add: "v0.0.422 added personal hooks and startup hooks."

Step 4 — Re-evaluate coverage delta. Claude Code has 14+ documented hook events (per the Claude Code changelog); Copilot CLI now has preToolUse, agentStop, subagentStop, personal hooks, and startup hooks (~5 confirmed). The coverage gap remains but is narrowing.

No Status column change required. The detail section is updated and the gap delta is recorded.

# Automate step 1: fetch latest Copilot CLI release and print entries
gh api repos/github/copilot-cli/releases/latest --jq '.body' | head -40

Related

• Cross-Tool Translation: Learning from Multiple AI Assistants
• CLI → IDE → GitHub: Context Ladder
• Claude Code /batch and Worktrees
• Worktree Isolation
• Hooks Lifecycle Events
• PostToolUse Hooks: Auto-Formatting on Every File Edit
• Skill Library Evolution
• Copilot CLI: Agentic Workflows

Last reviewed: 2026-05-27