@hasna/connectors 0.2.5 → 0.2.7

package/connectors/connect-cloudflare/CLAUDE.md

@@ -47,6 +47,27 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### New Workers REST API (Beta, Sep 2025)
+New resource-oriented REST API with cleaner separation:
+- `POST /workers/beta/workers` — create Worker (no code needed)
+- `POST /workers/beta/workers/{id}/versions` — upload code as version
+- `POST /scripts/{name}/deployments` — deploy a version
+
+Workers and Versions use new `/workers/` beta endpoints; Deployments remain on `/scripts/` endpoint.
+
+### Workers AI — New Capabilities (2026)
+- **Markdown Conversion** (`env.AI.toMarkdown()`): PDF, HTML, images with `conversionOptions` (CSS selectors, language)
+- **Real-time Transcription**: 10 languages via Deepgram Nova-3 on Workers AI in RealtimeKit
+- REST: `POST /accounts/{id}/ai/tomarkdown`
+
+### Browser Rendering — Rate Limit Increase (Mar 2026)
+Workers Paid plans: 3 req/s → **10 req/s** (600/min). Endpoints: `/links`, `/json`, `/scrape`, `/snapshot`, `/markdown`, `/pdf`, `/screenshot`, `/content`
+
+### Sandboxes — Real-time File Watching (Mar 2026)
+`sandbox.watch(path, options)` — SSE stream backed by inotify. Events: `create`, `modify`, `delete`, `move`.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-discord/CLAUDE.md

@@ -47,6 +47,25 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Current API Version: v10 (unchanged)
+
+### New in Oct 2025 (Developer Newsletter)
+- **Bot profiles per server**: Bots can now set unique profiles per server
+- **New UI components**: Label and TextDisplay components in modals; all select menus now work in modals
+- **New Social SDK webhooks**: `LOBBY_MESSAGE_*` and `GAME_DIRECT_MESSAGE_*` events for moderation
+- **Server-side Provisional Account API**: New endpoint for creating provisional accounts in games
+- **Rate limit**: New rate limit when requesting all guild members via gateway
+
+### Deprecations (Oct 2025)
+- Non-E2EE voice calls being phased out
+- Deprecated voice and video transport modes — update implementations
+
+### Bug Fixes (Mar 2026)
+- Escaped mentions `\@everyone` no longer trigger mentions — backend now properly respects escape characters
+- Desktop app p50 TTI reduced by 11.8% via API payload ordering optimization
+
 ## Authentication
 
 Bearer Token authentication. Credentials can be set via:

package/connectors/connect-docker/CLAUDE.md

@@ -47,6 +47,14 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Notes (2026)
+
+Docker Hub REST API: `https://hub.docker.com/v2/`
+Docker Engine API: `https://unix:///var/run/docker.sock/v1.xx/` (local socket)
+Current Engine API version: v1.47+ (Docker Engine 27+)
+
+Auth: Docker Hub uses JWT tokens (POST /users/login). Docker Engine uses unix socket or TCP.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-e2b/CLAUDE.md

@@ -47,6 +47,27 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2024-2025)
+
+### SDKs
+- Python: `e2b` or `e2b-code-interpreter`
+- JavaScript/TypeScript: `@e2b/code-interpreter`
+
+### Sandbox Lifecycle
+- Start: ~150ms spin-up time
+- Pause/resume: data persistence across sessions
+- `download_url`/`downloadUrl` method on sandbox
+- Env vars configurable on `sandbox.create()`
+
+### Desktop Sandbox (Dec 2024)
+Virtual desktop for LLM computer use. Isolated VMs with full desktop environment.
+
+### Interactive Charts (Oct 2024)
+Automatic chart detection from Matplotlib code via `runCode()`/`run_code()`. Returns chart data to client for custom rendering.
+
+### Multi-language Support
+Python, JavaScript/TypeScript, and other languages via code interpreter.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-elevenlabs/CLAUDE.md

@@ -47,6 +47,30 @@ src/
 └── index.ts       # Library exports
 ```
 
+## Models (2026)
+
+| Model | Description |
+|-------|-------------|
+| `eleven_v3` | **Eleven v3** — GA since Feb 2026. Most expressive TTS. Audio tags, multi-speaker dialogue, 70+ languages. 68% error reduction on numbers/symbols. |
+| `eleven_multilingual_v2` | Multilingual v2 — stable, recommended for real-time/conversational use |
+| `eleven_turbo_v2_5` | Turbo v2.5 — lowest latency, real-time |
+| `eleven_flash_v2_5` | Flash v2.5 — fastest, lowest cost |
+
+### Eleven v3 Features (GA Feb 2026)
+- **Audio tags**: `[excited]`, `[whispers]`, `[sighs]`, `[laughs]` inline in text
+- **Multi-speaker dialogue**: New `POST /v1/text-to-dialogue` endpoint
+- **70+ languages** with improved accuracy (99% error reduction for phone numbers, chemical formulas)
+- Requires more prompt engineering than v2 — not recommended for real-time/low-latency use cases
+
+### Text to Dialogue API (new endpoint)
+```json
+POST /v1/text-to-dialogue
+[
+  {"speaker_id": "voice-id-1", "text": "[cheerfully] Hello there!"},
+  {"speaker_id": "voice-id-2", "text": "Hey, how are you?"}
+]
+```
+
 ## Authentication
 
 API Key (Header) authentication. Credentials can be set via:

package/connectors/connect-exa/CLAUDE.md

@@ -47,6 +47,41 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Company Search (Jan 2026)
+New fine-tuned retrieval model with entity matching for company queries. Use `type="auto"` + `category="company"`.
+
+```bash
+curl -X POST https://api.exa.ai/search \
+  -H "x-api-key: $EXA_API_KEY" \
+  -d '{"query": "fintech companies in Switzerland", "type": "auto", "category": "company", "numResults": 10}'
+```
+
+Results now include **structured entity data** (`entities[]`) with:
+- `name`, `foundedYear`, `description`, `workforce.total`
+- `headquarters` (address, city, country)
+- `financials` (fundingTotal, fundingLatestRound)
+- `webTraffic.total`
+
+Supports: semantic queries, composite queries, funding queries, attribute filtering, named lookups.
+
+### Endpoints Overview (2026)
+| Endpoint | Description |
+|----------|-------------|
+| `POST /search` | Neural/keyword search (`type`: auto, neural, keyword) |
+| `POST /search` with `type=deep` | Multi-query deep search with structured answers |
+| `POST /findSimilar` | Find pages similar to a given URL |
+| `POST /contents` | Extract text, summaries, highlights from URLs |
+| `POST /answer` | Direct answers with citations |
+| `POST /research` | Async agent-style deep research |
+
+### Search Types
+- `neural` — semantic/embedding-based
+- `keyword` — fast keyword matching
+- `auto` — automatically selects best type
+- `deep` — multi-query with structured answers (slower, better for complex queries)
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-figma/CLAUDE.md

@@ -47,6 +47,40 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Variables API — Extended Collections (Nov 2025)
+Extended collections allow variants of existing variable collections with overrides while maintaining a single source of truth.
+
+New fields on `VariableCollection`:
+- `isExtension: boolean`
+- `parentVariableCollectionId: string` — only when isExtension is true
+- `rootVariableCollectionId: string`
+- `inheritedVariableIds: string[]`
+- `localVariableIds: string[]`
+- `variableOverrides: Record<variableId, Record<modeId, value>>`
+
+New `POST /v1/files/:key/variables` param: `initialModeIdToParentModeIdMapping`
+
+### Node Properties (Jan 2026)
+New node properties:
+- `transformModifiers` on TRANSFORM_GROUP nodes
+- `textPathStartData` on TEXT_PATH nodes
+- `variableWidthPoints` — variable width stroke profiles
+- `complexStrokeProperties` — brush and dynamic strokes
+
+### Rate Limits (Nov 2025)
+Published rate limits now enforced. Check [rate limits docs](https://developers.figma.com/docs/rest-api/rate-limits) for current values.
+
+### Figma MCP Server (beta, Sep 2025)
+Figma now has an official MCP Server in beta for AI agent access to design files.
+
+### OAuth App Publishing (Nov 2025)
+All OAuth apps must complete the new publishing flow via the Apps page. PAT generation via docs is temporarily removed — use the Postman workspace for testing.
+
+### Discovery API (Jan 2026)
+For Enterprise plans with Governance+: can now retrieve text of user-submitted AI prompts from Figma AI tools.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-firecrawl/CLAUDE.md

@@ -47,6 +47,33 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Endpoints (v1)
+| Endpoint | Description |
+|----------|-------------|
+| `POST /v1/scrape` | Scrape a single URL |
+| `POST /v1/crawl` | Crawl a website |
+| `POST /v1/map` | Map a website's URLs |
+| `POST /v1/search` | Search the web and scrape results (Jun 2025) |
+| `POST /v1/extract` | Extract structured data (v2: pagination, search, FIRE-1) |
+| `POST /v1/agent` | Agentic data gathering (Dec 2025) |
+| `POST /v1/batch/scrape` | Parallel agent batch processing |
+
+### New in 2026
+- **Browser Sandbox** (Feb 2026) — Fully managed isolated browser for agents. Zero config, pre-loaded tools.
+- **PDF Parser v2** (Feb 2026) — 3x faster Rust-based parser. Three auto-adapting modes (clean text, scanned, complex).
+- **Parallel Agents** (Jan 2026) — Batch `/agent` queries in spreadsheet/JSON format with streaming.
+- **Branding Format v2** (Feb 2026) — Improved logo extraction for AI agents.
+- **Spark 1 Pro/Mini models** (Jan 2026) — Flexible model selection for `/agent`. Mini is 60% cheaper.
+- **Claude Code plugin** (Feb 2026) — Official Firecrawl skill for Claude Code.
+
+### /extract v2 (Apr 2025)
+Added: pagination, FIRE-1 intelligent interaction, built-in search integration.
+
+### FIRE-1 Agent (Apr 2025)
+New AI agent that intelligently navigates and interacts with web pages for complex scraping.
+
 ## Authentication
 
 Bearer Token authentication. Credentials can be set via:

package/connectors/connect-github/CLAUDE.md

@@ -47,12 +47,42 @@ src/
 └── index.ts       # Library exports
 ```
 
-## Authentication
+## Authentication (2025+)
 
-Bearer Token authentication. Credentials can be set via:
+Bearer Token authentication. Use **fine-grained PATs** (GA since March 2025) for all new projects — they are now the recommended token type.
+
+Credentials can be set via:
 - Environment variable (see below)
 - Profile configuration: `connect-github config set-key <key>`
 
+### Token Types
+
+| Type | Recommendation |
+|------|---------------|
+| Fine-grained PAT | **Preferred** — GA since March 2025, granular permissions, audit-log `token_id` tracking |
+| Classic PAT | Legacy — still works, but over-permissioned |
+
+Fine-grained PAT gaps (not yet supported): Packages API, Checks API, internal repos outside target org, multi-org access.
+
+### Required Scopes (fine-grained PAT)
+
+- `contents:read/write` — repos, files
+- `issues:read/write` — issues, comments
+- `pull_requests:read/write` — PRs, reviews
+- `metadata:read` — required for all repo access
+- `models:read` — if using GitHub Models API (required since March 29 2025)
+
+### API Version
+
+Current: `X-GitHub-Api-Version: 2022-11-28` (unchanged)
+
+### GitHub MCP Server
+
+The Remote GitHub MCP Server is GA since September 2025. Add to your MCP config:
+```json
+{ "mcpServers": { "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"] } } }
+```
+
 
 ## Key Patterns
 

package/connectors/connect-gmail/CLAUDE.md

@@ -47,6 +47,28 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### OAuth Enforcement (May 2025)
+Google fully enforced OAuth 2.0 for all Gmail access as of May 2025. Less Secure Apps / password-based auth is permanently disabled. All clients MUST use OAuth 2.0 token-based authentication.
+
+### Scope Classification (2026)
+| Scope | Classification | Notes |
+|-------|---------------|-------|
+| `https://mail.google.com/` | Restricted | Full access |
+| `https://www.googleapis.com/auth/gmail.modify` | Restricted | Read + modify |
+| `https://www.googleapis.com/auth/gmail.readonly` | Restricted | Read only |
+| `https://www.googleapis.com/auth/gmail.send` | Sensitive | Send only |
+| `https://www.googleapis.com/auth/gmail.metadata` | Non-sensitive | Headers only |
+
+Restricted scopes require Google's restricted scope verification process including a third-party security assessment. For send-only use cases, prefer `gmail.send` to avoid stricter compliance requirements.
+
+### Gmail Postmaster Tools API v2 (GA Feb 2026)
+New: `POST /v2/domains/{domain}:queryDomainStats` — flexible querying of domain stats including compliance status. Supports batch operations. The v2beta endpoint was available Dec 2025.
+
+### Deal Cards in Promotions Tab (Sep 2025)
+New annotation support for creating Deal Cards. See Gmail annotations documentation.
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-google/CLAUDE.md

@@ -47,6 +47,17 @@ src/
 └── index.ts       # Library exports
 ```
 
+## Auth Notes (2026)
+
+Google OAuth 2.0 — use Application Default Credentials or service account JSON.
+
+### Maps Platform Client IDs Deprecated (May 2025+)
+Maps platform client IDs deprecated May 26, 2025, cannot be used after May 31, 2026.
+- In Apps Script: `setAuthentication(clientId, signingKey)` → deprecated Jun 2026
+- Use `setAuthenticationByKey(apiKey)` or `setAuthenticationByKey(apiKey, signingKey)` instead
+
+OAuth scopes are unchanged. userinfo endpoint: `https://www.googleapis.com/oauth2/v3/userinfo`
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-googlecalendar/CLAUDE.md

@@ -47,6 +47,15 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2026)
+
+### Google Meet Conference Guidance (Feb 2026)
+Do NOT reuse Google Meet codes across different calendar events — causes access issues and exposes meeting details to unintended users.
+Always generate a new conference using the `createRequest` field for every new event:
+```json
+"conferenceData": { "createRequest": { "requestId": "unique-id", "conferenceSolutionKey": { "type": "hangoutsMeet" } } }
+```
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-googlecloud/CLAUDE.md

@@ -47,6 +47,18 @@ src/
 └── index.ts       # Library exports
 ```
 
+## Auth Notes (2026)
+
+### Recommended Auth Methods
+1. **Application Default Credentials (ADC)** — `GOOGLE_APPLICATION_CREDENTIALS` env var pointing to service account JSON
+2. **Service Account Key** — JSON key file with email + private key
+3. **Workload Identity Federation** — keyless auth from AWS/Azure/on-prem (recommended for production)
+4. **OAuth 2.0 User Credentials** — for user-delegated access
+
+### Google AI / Gemini (separate from GCP)
+- Gemini API: `GEMINI_API_KEY` — simpler for Gemini models
+- Vertex AI: uses ADC or service account — more enterprise controls
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-googlecontacts/CLAUDE.md

@@ -47,6 +47,13 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Notes (2026)
+
+Use **Google People API** (not deprecated Contacts API v3).
+Base URL: `https://people.googleapis.com/v1/`
+Key endpoints: `/people/me`, `/people:listDirectoryPeople`, `/people:batchGet`
+Scopes: `contacts`, `contacts.readonly`, `directory.readonly`, `user.addresses.read`, `user.emails.read`
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-googledocs/CLAUDE.md

@@ -47,6 +47,13 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2026)
+
+### Gemini Features (Feb 2026)
+Audio summaries now available in Google Docs — Gemini-powered feature that generates audio overviews of documents.
+
+No breaking API changes to the Docs API v1 (stable).
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-googledrive/CLAUDE.md

@@ -47,6 +47,11 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2026)
+
+### Deprecation (Feb 2026)
+`enforceExpansiveAccess` query parameter is now deprecated for all methods in the permissions resource (v2 and v3). To restrict item access, use **folders with limited access** setting instead.
+
 ## Authentication
 
 OAuth authentication (Google OAuth2). Credentials can be set via:

package/connectors/connect-googlegemini/CLAUDE.md

@@ -47,6 +47,24 @@ src/
 └── index.ts       # Library exports
 ```
 
+## Models (2026)
+
+| Model | Description |
+|-------|-------------|
+| `gemini-3.1-pro-preview` | Latest (Feb 2026), #1 on 13/16 benchmarks, 1M ctx — **use for complex tasks** |
+| `gemini-3.1-pro-preview-customtools` | Agent workflows with custom tools (bash, code, multi-step) |
+| `gemini-3.1-flash-lite` | Fast, budget-friendly (preview) |
+| `gemini-3-pro-preview` | Gemini 3 Pro |
+| `gemini-3-flash-preview` | Frontier performance at fraction of cost |
+| `gemini-2.5-pro` | GA, advanced reasoning (discontinues Jun 17 2026) |
+| `gemini-2.5-flash` | Best price-performance, GA |
+| `gemini-2.5-flash-lite` | Fastest, most budget-friendly |
+| `gemini-2.5-flash-live-preview` | Real-time audio/video Live API |
+
+For image gen (Nano Banana): `gemini-2.5-flash-preview-image-generation`
+For video gen (Veo): `veo-3.1-generate-preview`
+For TTS: `gemini-2.5-flash-preview-tts`, `gemini-2.5-pro-preview-tts`
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-googlegemini/src/types/index.ts

@@ -2,19 +2,26 @@
 
 // ============ Models ============
 export type GeminiModel =
-  // Text Generation Models
+  // Gemini 3.1 (Feb 2026, latest)
+  | 'gemini-3.1-pro-preview'            // Most advanced, tops 13/16 benchmarks
+  | 'gemini-3.1-pro-preview-customtools' // Agent/tool-use optimized variant
+  | 'gemini-3.1-flash-lite'             // Fast, cost-efficient (preview)
+  // Gemini 3 (2025-2026)
   | 'gemini-3-pro-preview'
   | 'gemini-3-flash-preview'
+  // Gemini 2.5 (GA)
   | 'gemini-2.5-pro'
   | 'gemini-2.5-flash'
   | 'gemini-2.5-flash-lite'
-  // Image Generation Models (Nano Banana)
+  // Image Generation (Nano Banana)
   | 'gemini-2.5-flash-preview-image-generation'
   | 'gemini-3-pro-image-preview'
   // TTS Models
   | 'gemini-2.5-flash-preview-tts'
   | 'gemini-2.5-pro-preview-tts'
-  // Embedding Model
+  // Live API (real-time audio/video)
+  | 'gemini-2.5-flash-live-preview'
+  // Embedding
   | 'gemini-embedding-001'
   // Legacy
   | 'gemini-2.0-flash'

package/connectors/connect-googlemaps/CLAUDE.md

@@ -47,6 +47,21 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Places API v2 (New)
+New Places API (v2) is preferred over legacy Places API (v1). Use `https://places.googleapis.com/v1/places` endpoints.
+
+### Client IDs Deprecated (May 2025 → May 2026)
+Maps Platform client IDs deprecated May 26, 2025. Cannot be used after May 31, 2026.
+Use API key authentication exclusively.
+
+### Current APIs
+- Places API (New): `https://places.googleapis.com/v1/`
+- Routes API: `https://routes.googleapis.com/`
+- Geocoding API: `https://maps.googleapis.com/maps/api/geocode/json`
+- Maps JavaScript API, Street View, Distance Matrix, etc.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-googlesheets/CLAUDE.md

@@ -47,6 +47,16 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025)
+
+### Tables API (Apr 2025)
+Create, modify, and take basic actions on Tables in Google Sheets:
+- `addTable` request in `batchUpdate`
+- `Table` object in Sheets API
+
+### Smart Chips (Jun 2025)
+Create, update, and read Google Drive and People chips via the Sheets API.
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-googletasks/CLAUDE.md

@@ -47,6 +47,12 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Notes (2026)
+
+Google Tasks API v1 is stable. Base URL: `https://tasks.googleapis.com/tasks/v1/`
+Key endpoints: `/lists`, `/lists/{list}/tasks`, `/users/@me/lists`
+Required scope: `https://www.googleapis.com/auth/tasks`
+
 ## Authentication
 
 OAuth authentication. Credentials can be set via:

package/connectors/connect-hedra/CLAUDE.md

@@ -47,6 +47,21 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API (Full Platform, Feb 2026)
+
+Full Hedra API launched Feb 2026. All models available programmatically.
+
+New API endpoints (unified):
+- `GET /models` — list models
+- `POST /assets/generate` — generate image/video/avatar
+- `GET /assets/{id}` — check status
+- `GET /credits` — credit balance
+
+Legacy Character-2 API: `POST /v1/characters` (still works)
+Auth: `X-API-Key` header
+
+Capabilities: image generation, video generation, avatar generation, Live Avatar (real-time).
+
 ## Authentication
 
 Bearer Token authentication. Credentials can be set via:

package/connectors/connect-heygen/CLAUDE.md

@@ -47,6 +47,20 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Avatar IV Endpoint (Jan 2026)
+```
+POST /v2/video/av4/generate
+```
+Photorealistic full-body avatar videos. Speed/quality mode selection.
+
+### Template API Variables (Aug 2025)
+AI Studio templates now support API creation with variable customization: avatars, text, elements, scene scripts.
+
+### Digital Twin — Avatar IV (Aug 2025)
+Full-body digital twins with gestures and emotion.
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-huggingface/CLAUDE.md

@@ -47,6 +47,26 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Updates (2025-2026)
+
+### Serverless Inference API — Overhauled (2026)
+The Serverless Inference API is being completely overhauled. Key changes:
+
+New router endpoint pattern:
+```
+https://router.huggingface.co/hf-inference/models/{model}/v1/chat/completions
+```
+
+For chat completions (OpenAI-compatible):
+```bash
+curl -X POST https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.1-8B-Instruct/v1/chat/completions \
+  -H "Authorization: Bearer $HF_TOKEN" \
+  -H "Content-Type: application/json" \
+  -d '{"model": "meta-llama/Llama-3.1-8B-Instruct", "messages": [{"role": "user", "content": "Hello"}]}'
+```
+
+Note: Error responses may return HTML 503 during migration period — handle non-JSON responses gracefully.
+
 ## Authentication
 
 Bearer Token authentication. Credentials can be set via:

package/connectors/connect-icons8/CLAUDE.md

@@ -47,6 +47,12 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Notes (2026)
+
+Icons8 API: `https://api.icons8.com/api/iconsets/v5/`
+Key endpoints: `/search`, `/icon`, `/categories`, `/styles`
+Auth: `token` query parameter
+
 ## Authentication
 
 API Key authentication. Credentials can be set via:

package/connectors/connect-maropost/CLAUDE.md

@@ -47,6 +47,12 @@ src/
 └── index.ts       # Library exports
 ```
 
+## API Notes (2026)
+
+Maropost Marketing Cloud API: `https://api.maropost.com/accounts/{account_id}/`
+Auth: `auth_token` query parameter or Bearer token header.
+Key resources: contacts, campaigns, tags, lists, journeys, reports.
+
 ## Authentication
 
 Bearer Token authentication. Credentials can be set via:

package/connectors/connect-mercury/CLAUDE.md

@@ -47,6 +47,20 @@ src/
 └── index.ts       # Library exports
 ```
 
+## New Endpoints (2025-2026)
+
+Mercury has been rapidly expanding their API. Recent additions:
+- **Webhooks** — real-time HTTP notifications for account events (transactions, etc.)
+- **Internal Transfers** — `POST /transfer` — move money between accounts (requires `SendMoney` write scope)
+- **Users API** — `GET /api/v1/users` — retrieve organization users
+- **Account Balance Events** — webhooks/events for balance change notifications
+- **SAFE API** — 3 new endpoints for reading SAFE (Simple Agreement for Future Equity) data
+- **Treasury API statements** — programmatic statement retrieval
+- **Attachment uploads** — upload attachments to transactions and recipients
+
+Auth: Basic auth (API token as username, empty password) or Bearer token.
+Token tiers: Read-only, Read-write (requires IP whitelist), Custom (scoped).
+
 ## Authentication
 
 API Key authentication. Credentials can be set via: