- Add --bare flag to `ccs auth create` to skip shared symlinks
(commands/, skills/, agents/, settings.json). Bare profiles are
clean instances without ClaudeKit or other global extensions.
`ccs sync` respects bare flag and skips re-linking.
- Add MCP server sync from global ~/.claude.json to instances.
Claude Code stores MCP config in ~/.claude.json (not settings.json),
which was invisible to CCS profiles using CLAUDE_CONFIG_DIR isolation.
Selective copy of mcpServers key only (not OAuth sessions/caches).
`ccs sync` refreshes MCP servers across all non-bare instances.
Closes#691Closes#692
Fixes profile creation for local providers (llamacpp, ollama) where
apiKey is not required but server validation requires a truthy value.
Changes:
- Set apiKey to preset.apiKeyPlaceholder when selecting a preset
- Use preset.apiKeyPlaceholder as fallback in onSubmit
- Ensures sentinel values (e.g., "llamacpp", "ollama") are sent
Fixes user-reported error when creating llamacpp profile via dashboard.
- Add llamacpp provider preset to provider-presets.ts
- Add base-llamacpp.settings.json configuration
- Update help-command.ts with ccs llamacpp documentation
- Update UI provider presets for dashboard parity
llama.cpp is a popular C++ inference engine for LLaMA models.
This complements existing Ollama support for local inference.
Usage: ccs llamacpp (requires llama.cpp server on port 8080)
Closes#690
Support native Anthropic API keys (sk-ant-...) via `ccs api create --preset anthropic`.
Profiles use ANTHROPIC_API_KEY env var without BASE_URL, letting Claude CLI
authenticate natively with x-api-key header instead of proxy Bearer token.
- Add anthropic preset to provider catalog
- Branch profile-writer env structure (native vs proxy)
- Add validateAnthropicKey() preflight with x-api-key header
- Auto-detect api.anthropic.com URL in interactive flow
- Fix hasApiKey/isConfigured to recognize ANTHROPIC_API_KEY
- Update delegation-validator for native mode profiles
- Add Droid target API key fallback (ANTHROPIC_API_KEY)
Closes#688
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
largeFileCountOver350Loc is a vanity metric that blocks releases
for borderline cases (359-372 LOC files). Keep processExit and
synchronousFs gates which track real anti-patterns. LOC still
collected/reported but no longer gated.
largeFileCountOver350Loc is a vanity metric that blocks releases
for borderline cases (359-372 LOC files). Keep processExit and
synchronousFs gates which track real anti-patterns. LOC still
collected/reported but no longer gated.
When models have thinking enabled (e.g. codex with thinking: auto), the
API returns content as [{ type: 'thinking', thinking: '...' }, { type:
'text', text: '...' }]. The hook reads content[0].text which is
undefined for thinking blocks, causing 'No text content in response'
errors that block the Read tool (exit 2).
Also fix cross-provider model fallback: getModelsToTry() unconditionally
appends DEFAULT_MODEL (gemini-2.5-flash) as retry, which 502s on
non-gemini provider routes (e.g. codex). Only use DEFAULT_MODEL when no
provider-specific model is configured.
Fixes: content[0].text -> content.find(b => b.type === 'text')
Fixes: cross-provider retry causing 'unknown provider' 502
Ref: #511
