feat(providers): add GitHub Copilot community provider

.archon/workflows/test-workflows/e2e-copilot-abort.yaml

@@ -0,0 +1,14 @@
+# E2E manual abort test — GitHub Copilot community provider
+# Verifies: Ctrl-C propagates through the bridge to session.abort() and
+#   sendAndWait unwinds cleanly without dangling listeners.
+# Manual: start, wait for streaming to begin, press Ctrl-C. Not for CI.
+name: e2e-copilot-abort
+description: 'Manual test: start, then Ctrl-C. Verifies abort wiring.'
+provider: copilot
+model: gpt-5-mini
+
+nodes:
+  - id: long
+    prompt: 'Count slowly from 1 to 200, one number per line, with a brief phrase after each number explaining its mathematical significance. Do not skip any numbers.'
+    effort: low
+    idle_timeout: 120000

.archon/workflows/test-workflows/e2e-copilot-all-nodes-smoke.yaml

@@ -0,0 +1,147 @@
+# E2E smoke test — Copilot provider, every CI-compatible node type
+# Covers: prompt, command, loop (AI node types) + bash, script bun/uv
+#   (deterministic node types) + depends_on / when / trigger_rule / $nodeId.output
+#   (DAG features) + Copilot-specific options: effort, allowed_tools,
+#   output_format (best-effort JSON via prompt augment + 2-tier parser).
+# Skipped: `approval:` — pauses for human input, incompatible with CI.
+# Auth: `gh auth login` OR `COPILOT_GITHUB_TOKEN`.
+#   To use `GH_TOKEN` / `GITHUB_TOKEN`, also set `assistantConfig.useLoggedInUser: false`.
+#   Requires an active GitHub Copilot subscription.
+name: e2e-copilot-all-nodes-smoke
+description: 'Copilot provider smoke across every CI-compatible node type plus Copilot-specific options.'
+provider: copilot
+model: gpt-5-mini
+
+nodes:
+  # ─── AI node types ──────────────────────────────────────────────────────
+
+  # 1. prompt: inline prompt + effort + allowed_tools (no tool calls).
+  #    Verifies reasoningEffort and availableTools=[] reach the SDK.
+  - id: prompt-node
+    prompt: "Reply with exactly the single word 'ok' and nothing else."
+    allowed_tools: []
+    effort: low
+    idle_timeout: 30000
+
+  # 2. command: named command file (.archon/commands/e2e-echo-command.md).
+  #    The command echoes back $ARGUMENTS (the workflow invocation message).
+  - id: command-node
+    command: e2e-echo-command
+    allowed_tools: []
+    idle_timeout: 30000
+
+  # 3. loop: iterative AI prompt until completion signal.
+  #    Bounded by max_iterations: 2 so a misbehaving model can't hang CI.
+  - id: loop-node
+    loop:
+      prompt: "Reply with exactly 'DONE' and nothing else."
+      until: 'DONE'
+      max_iterations: 2
+    allowed_tools: []
+    effort: low
+    idle_timeout: 60000
+
+  # 4. output_format: Copilot's best-effort structured output path
+  #    (prompt augmented with schema + 2-tier JSON parser on result text).
+  #    Unique to Copilot/Pi vs. Claude/Codex native JSON mode — only an
+  #    E2E test catches "real model drifted around the schema".
+  - id: structured-node
+    prompt: |
+      Return a JSON object with two fields, no fences and no prose:
+        - "status": always "ok" (string)
+        - "value":  always 42 (number)
+    allowed_tools: []
+    effort: low
+    idle_timeout: 30000
+    output_format:
+      type: object
+      properties:
+        status:
+          type: string
+        value:
+          type: number
+      required: [status, value]
+
+  # ─── Deterministic node types (no AI) ───────────────────────────────────
+
+  # 5. bash: shell script with JSON output (enables $nodeId.output.status
+  #    dot-access downstream).
+  - id: bash-json-node
+    bash: 'echo ''{"status":"ok"}'''
+
+  # 6. script: bun (TypeScript/JavaScript runtime)
+  - id: script-bun-node
+    script: echo-args
+    runtime: bun
+    timeout: 30000
+
+  # 7. script: uv (Python runtime)
+  - id: script-python-node
+    script: echo-py
+    runtime: uv
+    timeout: 30000
+
+  # ─── DAG features ───────────────────────────────────────────────────────
+
+  # 8. depends_on + $nodeId.output substitution
+  - id: downstream
+    bash: "echo 'downstream got: $prompt-node.output'"
+    depends_on: [prompt-node]
+
+  # 9. when: conditional (JSON dot-access on bash JSON output)
+  - id: gated
+    bash: "echo 'gated-ok'"
+    depends_on: [bash-json-node]
+    when: "$bash-json-node.output.status == 'ok'"
+
+  # 10. when: conditional on AI structured output (proves output_format
+  #     parsed and dot-access works on the resulting object).
+  - id: structured-check
+    bash: "echo \"structured.status=$structured-node.output.status\""
+    depends_on: [structured-node]
+    when: "$structured-node.output.status == 'ok'"
+
+  # 11. trigger_rule: merge multiple deps (all_success semantics)
+  - id: merge
+    bash: "echo 'merge-ok'"
+    depends_on:
+      [downstream, gated, structured-check, script-bun-node, script-python-node]
+    trigger_rule: all_success
+
+  # ─── Final assertion ────────────────────────────────────────────────────
+
+  # 12. Verify every upstream node produced non-empty output, including
+  #     dot-access on the structured-output node (proves output_format
+  #     parsed and downstream consumers can index into it).
+  #     Note: value-equality on string fields is avoided on purpose —
+  #     shellQuote() wraps strings in literal single quotes, so a literal
+  #     `[ "$x" != "ok" ]` would always fail. Non-emptiness is the right
+  #     bar for a smoke; the `when:` gate on structured-check already
+  #     proved the value matched 'ok' to reach this node.
+  - id: assert
+    bash: |
+      fail=0
+      check() {
+        local name="$1"
+        local value="$2"
+        if [ -z "$value" ]; then
+          echo "FAIL: $name produced empty output"
+          fail=1
+        fi
+      }
+      check prompt-node       "$prompt-node.output"
+      check command-node      "$command-node.output"
+      check loop-node         "$loop-node.output"
+      check bash-json-node    "$bash-json-node.output"
+      check script-bun-node   "$script-bun-node.output"
+      check script-python-node "$script-python-node.output"
+      check downstream        "$downstream.output"
+      check gated             "$gated.output"
+      check merge             "$merge.output"
+      check structured.status "$structured-node.output.status"
+      check structured.value  "$structured-node.output.value"
+
+      if [ "$fail" -eq 1 ]; then exit 1; fi
+      echo "PASS: all node types + structured output verified"
+    depends_on: [merge, loop-node, command-node]
+    trigger_rule: all_success

.env.example

@@ -38,6 +38,18 @@ CODEX_REFRESH_TOKEN=
 CODEX_ACCOUNT_ID=
 # CODEX_BIN_PATH=  # Optional: path to Codex native binary (binary builds only)
 
+# GitHub Copilot (community provider — @github/copilot-sdk)
+# Requires an active GitHub Copilot subscription. By default, Archon uses
+# the credentials you configured via the Copilot CLI (`copilot login`).
+# Generic GH_TOKEN / GITHUB_TOKEN (declared below) are intentionally NOT
+# picked up — classic PATs lack Copilot entitlement and would fail. To
+# opt back into env-token auth, set `useLoggedInUser: false` in
+# `.archon/config.yaml`. Setting COPILOT_GITHUB_TOKEN is treated as
+# explicit Copilot intent and always wins.
+#
+# COPILOT_GITHUB_TOKEN=        # Copilot-scoped PAT (always wins when set)
+# COPILOT_BIN_PATH=            # Optional: path to Copilot CLI binary (binary builds only)
+
 # Pi (community provider — @mariozechner/pi-coding-agent)
 # One adapter, ~20 LLM backends. Archon's Pi adapter picks up credentials
 # you've already configured via the Pi CLI (`pi /login` writes to
@@ -57,7 +69,7 @@ CODEX_ACCOUNT_ID=
 # OPENROUTER_API_KEY=          # Pi provider id: openrouter
 # HUGGINGFACE_API_KEY=         # Pi provider id: huggingface
 
-# Default AI Assistant (must match a registered provider, e.g. claude, codex, pi)
+# Default AI Assistant (must match a registered provider, e.g. claude, codex, copilot, pi)
 # Used for new conversations when no codebase specified — errors on unknown values
 DEFAULT_AI_ASSISTANT=claude
 

packages/core/src/config/config-loader.ts

@@ -99,6 +99,7 @@ const SAFE_ASSISTANT_FIELDS: Record<string, readonly string[]> = {
   // community providers — list each field we're confident is safe to
   // show in the web UI. Unknown providers fall through with no fields.
   pi: ['model'],
+  copilot: ['model'],
 };
 
 function toSafeAssistantDefaults(assistants: AssistantDefaults): SafeConfig['assistants'] {

packages/core/src/config/config-types.ts

@@ -16,13 +16,15 @@
 import type {
   ClaudeProviderDefaults,
   CodexProviderDefaults,
+  CopilotProviderDefaults,
   PiProviderDefaults,
   ProviderDefaultsMap,
 } from '@archon/providers/types';
 
 export type {
   ClaudeProviderDefaults,
   CodexProviderDefaults,
+  CopilotProviderDefaults,
   PiProviderDefaults,
   ProviderDefaultsMap,
 };

packages/docs-web/src/content/docs/getting-started/ai-assistants.md

@@ -1,6 +1,6 @@
 ---
 title: AI Assistants
-description: Configure Claude Code, Codex, and Pi as AI assistants for Archon.
+description: Configure Claude Code, Codex, GitHub Copilot, and Pi as AI assistants for Archon.
 category: getting-started
 area: clients
 audience: [user]
@@ -9,7 +9,7 @@ sidebar:
   order: 4
 ---
 
-You must configure **at least one** AI assistant. All three can be configured and mixed within workflows.
+You must configure **at least one** AI assistant. All four can be configured and mixed within workflows.
 
 ## Claude Code
 
@@ -389,6 +389,95 @@ Unsupported YAML fields trigger a visible warning from the dag-executor when the
 - [Adding a Community Provider](../contributing/adding-a-community-provider/) — the contributor-facing guide for extending Archon with your own provider.
 - [Pi on GitHub](https://github.com/badlogic/pi-mono) — upstream project.
 
+## GitHub Copilot (Community Provider)
+
+**Use a GitHub Copilot subscription inside Archon workflows.** Drives the Copilot CLI via `@github/copilot-sdk`, supporting OpenAI, Anthropic via BYOK, Gemini, and the other models Copilot exposes — switch between them with the `model` field.
+
+Copilot is registered as `builtIn: false` — like Pi, a bundled community provider rather than a core built-in.
+
+### Install
+
+For source installs (`bun run`), the SDK + its bundled CLI dependency come along with `bun install` — nothing extra to do.
+
+For compiled Archon binaries, install the Copilot CLI yourself and point Archon at it:
+
+```bash
+npm install -g @github/copilot
+```
+
+Then tell Archon where the binary lives (the resolver searches these in order):
+
+```ini
+# .env
+COPILOT_BIN_PATH=/absolute/path/to/copilot
+```
+
+```yaml
+# .archon/config.yaml
+assistants:
+  copilot:
+    copilotCliPath: /absolute/path/to/copilot
+```
+
+Or place the binary at `~/.archon/vendor/copilot/copilot` (POSIX) / `~/.archon/vendor/copilot/copilot.exe` (Windows) and the resolver picks it up automatically.
+
+### Authenticate
+
+By default, Copilot uses the credentials from your local `copilot login`. Generic `GH_TOKEN` / `GITHUB_TOKEN` env vars are **not** picked up automatically — classic GitHub PATs lack Copilot entitlement and would fail with a misleading SDK error. Auth precedence (highest to lowest):
+
+1. **`COPILOT_GITHUB_TOKEN`** (env) — always wins when set; treated as explicit Copilot intent
+2. **`useLoggedInUser: false`** in `.archon/config.yaml` — opts into env-token auth, including generic `GH_TOKEN` / `GITHUB_TOKEN`
+3. **`copilot login` credentials** — the default
+
+An active GitHub Copilot subscription is required for any of these to work.
+
+### Copilot Configuration Options
+
+You can configure Copilot's behavior in `.archon/config.yaml`:
+
+```yaml
+assistants:
+  copilot:
+    model: gpt-5-mini             # 'gpt-5', 'gpt-5-mini', 'claude-sonnet-4.5', 'auto', etc.
+    modelReasoningEffort: medium  # 'low' | 'medium' | 'high' | 'xhigh' | 'max' (alias for xhigh)
+    # configDir: /absolute/path/to/copilot-config
+    # enableConfigDiscovery: false  # only enable for trusted repos — bypasses Archon's workflow MCP/skill validation
+    # useLoggedInUser: false        # opt into env-token auth (GH_TOKEN / GITHUB_TOKEN); default uses `copilot login`
+    # logLevel: error               # 'none' | 'error' | 'warning' | 'info' | 'debug' | 'all'
+```
+
+Copilot accepts OpenAI models (`gpt-5`, `gpt-5-mini`), Anthropic via BYOK (`claude-sonnet-4.5`), Gemini, and more. When no model is configured, Archon passes `model: 'auto'` and Copilot picks.
+
+### Supported Archon Features
+
+| Feature | Support | Notes |
+|---|---|---|
+| Session resume | ✅ | Returns `sessionId`; reused on resume |
+| Reasoning control | ✅ | `effort:` / string `thinking:` → Copilot `reasoningEffort`; `max` maps to SDK `xhigh` |
+| System prompt override | ✅ | `systemPrompt:` |
+| Codebase env vars | ✅ | merged into the spawned Copilot CLI environment |
+| Tool restrictions | ✅ | `allowed_tools` → `availableTools`, `denied_tools` → `excludedTools` |
+| MCP servers | ✅ | `mcp: path/to/servers.json` → `SessionConfig.mcpServers` (env vars `$FOO` expanded; missing vars warned) |
+| Skills | ✅ | `skills: [name]` resolved from `.agents/skills/` or `.claude/skills/` (project or home) → `SessionConfig.skillDirectories` |
+| Structured output | ✅ | best-effort via prompt augmentation; unparseable output degrades to dag-executor's missing-output warning |
+| Sub-agents (`agents:`) | ✅ | `name`/`description`/`prompt`/`tools` → `SessionConfig.customAgents`; Claude-specific fields (`model`, `disallowedTools`, `skills`, `maxTurns`) warn per agent and are ignored |
+| Fork-session retry | ⚠️ | Copilot SDK has no fork API — when Archon requests a fork (on retry), we create a fresh session and emit a system-chunk warning |
+| Hooks | ❌ | Archon hooks ≠ Copilot's `SessionHooks` event vocabulary |
+| Fallback model | ❌ | not wired |
+| Cost control | ❌ | no cost-limit API |
+| Sandbox | ❌ | Copilot permissions surface is separate from Archon's sandbox model |
+
+### Set as Default (Optional)
+
+```ini
+DEFAULT_AI_ASSISTANT=copilot
+```
+
+### See also
+
+- [Adding a Community Provider](../contributing/adding-a-community-provider/) — the contributor-facing guide for extending Archon with your own provider.
+- [`@github/copilot-sdk`](https://www.npmjs.com/package/@github/copilot-sdk) — upstream SDK.
+
 ## How Assistant Selection Works
 
 - Assistant type is set per codebase via the `assistant` field in `.archon/config.yaml` or the `DEFAULT_AI_ASSISTANT` env var