v0.4.9

Fixed

• Deploy button targets the worker subdirectory (infra/worker) so Cloudflare detects the wrangler config and provisions D1 + KV correctly
• Worker-scoped deploy-form secrets with CHANGE_ME placeholders + in-form binding descriptions; runtime treats placeholders as unset
• Setup prompt: consent-first guided flow, cloud default with npx/local fallbacks, per-agent connector configs, redirect-mismatch self-recovery

v0.4.8

Changed

• npm metadata: expanded keywords for search discovery (mcp, model-context-protocol, content-creator, ai-agent, claude, chatgpt, gemini, cursor, opencode, …)
• Sharper package description

v0.4.7

Fixed

• npm package now includes README.md + LICENSE on npmjs.com
• Docker build: CI=true env var prevents pnpm TTY prompt in container

v0.4.6

Added

• Publish to GitHub Packages (npm.pkg.github.com) — visible on repo's Packages tab
• Docker build fix: corepack fallback for node:25-slim images

Changed

• README updated with scoped npm name @aryaminus/socials-mcp and package links

v0.4.5

Fixed

• Release smoke workflow: tarball glob pattern matches scoped package name aryaminus-socials-mcp-*.tgz

v0.4.3

Fixed

• npm publish: use npm config set instead of setup-node registry-url to avoid token conflict

v0.4.2

Fixed

• npm publish: hardcode .npmrc auth token in CI workflow (setup-node .npmrc generation was not passing token correctly)

v0.4.1

Added

• docs/SETUP-PROMPT.md: single magic prompt users paste into any AI agent (Claude, ChatGPT, Gemini, Codex, Cursor, DeepSeek, Qwen, Pi, opencode, Claude Desktop, Claude Code, or any MCP-compatible agent) for full end-to-end setup

Changed

• All docs now agent-agnostic — no longer claude.ai-specific; every agent listed equally (Claude, ChatGPT, Gemini, Codex, Cursor, DeepSeek, Qwen, Pi, opencode, Claude Desktop, Claude Code, Antigravity, Z.ai, and any MCP-compatible agent)
• README.md: one-prompt setup section prominent at top
• AGENTS.md: one-prompt setup reference, all agents listed in install table
• GET-STARTED-CLOUD.md: agent-agnostic connector instructions with per-agent table

Fixed

• npm publish: removed --registry flag that overrode .npmrc token from setup-node

v0.4.0

Added

• Zero-install cloud journey (docs/GET-STARTED-CLOUD.md): Deploy-to-Cloudflare button (root wrangler.jsonc auto-provisions Worker+D1+KV), self-service /setup checklist page served by the Worker (no CLI), in-chat platform connections (platform_oauth_url / platform_oauth_exchange — agent issues consent link, human pastes the redirect back; tokens exchanged server-side, stored encrypted)
• Platform-OAuth landing page (/oauth/*/done) with paste-back instructions
• Release pipeline aligned with controlkeel pattern: release-smoke.yml (builds + tests + skill bundles on version-bump commits), release.yml (tag-driven, waits for smoke, cosign + checksums, GitHub Release, npm publish with provenance)
• .env.example template with all environment variables documented
• script-review skill: search demand research, Good/Bad/Ugly summary, on-screen text cues with timestamps, mid-video re-hook suggestions, content-mix fit dimension, plain language rule

Changed

• README rewritten: concise, no redundant info, points to docs for details
• Removed Render deployment (render.yaml) and Smithery marketplace (smithery.yaml)
• Removed Render/Fly/Railway from provider comparisons in infra/worker/README.md
• CI (ci.yml) skips version-bump commits, concurrency groups, paths-ignore for docs
• apps/mcp/package.json: aligned @modelcontextprotocol/sdk to ^1.30.0, zod to ^3.25.0; fixed main/exports to point to dist-bundle/
• AGENTS.md: full setup reference (GitHub secrets, Cloudflare secrets, local env, npm publish), missing env vars added (SOCIALS_CSV_DIR, SOCIALS_OAUTH_PORT, SOCIALS_ALLOWED_ORIGINS)

Removed

• render.yaml (Render deployment)
• smithery.yaml (Smithery marketplace listing)