83 posts tagged with "AI Agents"

Supabase ships an official MCP server that gives your AI access to the full Supabase platform — Postgres databases, authentication, storage, edge functions, and project management. It's one of the more complete official MCP implementations, covering both development workflows and production operations.

This guide covers what the Supabase MCP server offers, how to set it up, and how to access it through MCPBundles alongside your other tools.

Linear has one of the best official MCP servers in the ecosystem. Built in partnership with Cloudflare and Anthropic, it runs as a remote hosted server at mcp.linear.app/mcp with OAuth 2.1 authentication and 25+ tools for managing your entire project workflow.

This guide covers what the Linear MCP server can do, how to set it up across different AI clients, and how to access it through MCPBundles for a hosted experience alongside your other tools.

There's no official PostgreSQL MCP server from the PostgreSQL Foundation — and there probably won't be, since PostgreSQL is an open-source project without a commercial entity pushing integrations. The community implementations that exist mostly stop at "run a query, get rows" — fine as a starting point, light on everything that actually shows up in real database work.

MCPBundles provides 20+ purpose-built tools that go far beyond raw SQL. Your AI explores schemas, profiles columns, analyzes index health, detects data quality issues, finds duplicates, explains query plans, and exports data — all without you writing a single SQL statement. And if you do want raw SQL, that's there too.

Most Discord server management is repetitive moderation and community work. Read every message in #support to find unanswered questions, then draft and post threaded replies. Scan #general for the day's key discussions and post a summary to #daily-digest. Create separate discussion threads for each agenda item in a pinned meeting note. React with checkmarks to every completed task message. Pin important announcements so they're easy to find.

Each of those is a 15-minute task in the Discord UI and a 30-second task as a chat message — if your AI agent can actually call the Discord API. This guide is the use-case version of "AI + Discord": what you ask, what the agent does, what comes back. The protocol underneath is MCP (Model Context Protocol), the bundle is /skills/discord on MCPBundles, but the framing here is workflow-first.

Most performance-marketing work in Google Ads is repetitive cognitive labour. Pull a search-term report, find the queries that wasted spend last week, write the negative-keyword list. Look at device performance, find that mobile CPC is up 40% with the same conversion rate, draft a bid adjustment. Spin up a campaign for next week's promo: budget, ad group, 15 keywords, an RSA with 11 headlines and 4 descriptions, all in PAUSED so nothing goes live by accident.

Each of those is a 20-minute task in the Google Ads UI and a 30-second task as a chat message — if your AI agent can actually call the Google Ads API. This guide is the use-case version of "AI + Google Ads": what you ask, what the agent does, what comes back. The protocol underneath is MCP (Model Context Protocol), the bundle is /skills/google-ads on MCPBundles, but the framing here is workflow-first.

Every other LinkedIn MCP server on GitHub is either a scraper that violates LinkedIn's Terms of Service or a thin wrapper around unofficial endpoints that can break at any time. Some use Patchright (a Playwright fork) to automate the browser. Others reverse-engineer private APIs. LinkedIn actively blocks these — and your account is at risk if you use them.

MCPBundles ships two LinkedIn MCP servers that work the way LinkedIn allows you to work:

• LinkedIn — built on LinkedIn's official REST API with OAuth 2.0. Manages company pages, publishes posts with images and carousels, engages with comments and reactions, runs ad campaigns, and tracks analytics.
• LinkedIn Sales Navigator — drives Sales Navigator search, profile lookups, InMail, and inbox as your connected LinkedIn account. Sign in once on a hosted page (no cookies, no extensions, no scraping); the bundle then runs SN-grade prospecting and outreach with built-in per-account rate caps so a careless agent can't get the seat banned.

You can connect either or both. The pages are the canonical product surfaces — keep this article open if you want the comparison; jump to the skill page when you're ready to use one.

Most small-business accounting is repetitive operational work. Create an invoice for last week's hours, send it to the customer, and record the payment when it hits the bank. Pull a P&L for the quarter, export an aged-receivables report, check which customers are 30+ days overdue. Batch-update item prices, void an invoice that was sent to the wrong contact, reconcile what changed this week across customers, invoices, bills, and payments.

Each of those is a 5–10 minute task in the QuickBooks UI and a 20-second task as a chat message — if your AI agent can actually call the QuickBooks API. This guide is the use-case version of "AI + QuickBooks": what you ask, what the agent does, what comes back. The protocol underneath is MCP (Model Context Protocol), the bundle is /skills/quickbooks on MCPBundles, but the framing here is workflow-first.

Stanford's Digital Economy Lab just published The Enterprise AI Playbook — a 116-page study of 51 successful enterprise AI deployments across 41 organizations, 9 industries, and 7 countries. The research team, led by Erik Brynjolfsson (one of the most-cited economists on technology), interviewed executives and project leads who deployed AI at scale and measured actual results.

The headline finding: the technology was never the hard part. In 77% of cases, the hardest challenges were invisible — change management, data quality, and process redesign. Not model selection. Not prompt engineering. Not which AI provider to use.

This post pulls out the findings that matter most for anyone building or buying AI tooling today.

The terminal is having its best year since the invention of cloud infrastructure.

Every major AI lab shipped a coding agent CLI. Every major SaaS company shipped or meaningfully updated a service CLI. And a new category is emerging — CLIs that connect the two, giving your coding agent access to production services without leaving the terminal.

We've been running MCPBundles for over a year — a platform where teams connect AI agents to production APIs. We built a CLI because we kept watching agents context-switch between writing code and needing to call Stripe, query a database, or check analytics. This guide covers everything worth installing in 2026, organized by what it actually does for you.

OpenClaw — the open-source personal AI assistant with over 340,000 GitHub stars — now supports MCP. That means it can connect to any MCP server and use external tools as first-class capabilities. Your personal AI can go from answering questions to actually doing things across your services.

We've been watching OpenClaw's growth since late 2025. It runs on your hardware, works with Claude, GPT-4, Gemini, DeepSeek, and local models through Ollama, and connects to 20+ messaging platforms — WhatsApp, Telegram, Discord, Slack, Signal, iMessage. It already had 3,200+ skills on ClawHub. What it didn't have was a standardized way to reach the services those skills talked about.

MCP changes that. And MCPBundles makes it trivial.

There's a war happening on Reddit right now, and it's getting heated.

On one side: developers who believe the Model Context Protocol is overengineered middleware — that AI agents should just call gh issue create and curl like any terminal user. On the other: engineers running MCP in production who say the skeptics will inevitably reinvent every feature MCP provides, just worse.

Both sides are partially right. But the debate itself is framed wrong.

I spent the last day using our MCPBundles CLI to search Reddit via MCP tools — browsing posts, pulling comment threads, analyzing arguments — all through authenticated MCP tool calls executed from the command line. The irony was not lost on me: I was using CLI to call MCP to read arguments about whether we need MCP or CLI.

The answer, as it turns out, is both. But not in the way most people think.

We ran into a problem a few weeks ago that none of our tools could solve. It wasn't a technical problem — the code was fine, the infra was fine. We just needed someone to go do a thing on a website. Sign up, click around, grab some information, paste it into a form. Repeat a bunch of times.

AI couldn't do it. The sites had captchas, email verification, multi-step flows. We tried browser automation and it broke immediately. We needed a person.

So we thought: what if our AI agent could just hire one?

The Glama directory lists 22,775 MCP servers as of May 2026. Most are weekend projects. Some are brilliant. BlueRock Security found 36.7% of public MCP servers carry SSRF vulnerabilities, 41% have no authentication at all, and only 8.5% use OAuth — so a "list of every MCP server" is not a useful list.

This guide is the opposite: the ~80 servers that real teams run in production, grouped by job. Four to five of them will cover 80% of what you ask your AI to do. We've been running MCPBundles for over a year — a platform where teams connect their AI agents to production APIs — and have tested, wrapped, and maintained MCP servers for hundreds of services. This is what we've learned about which ones are worth your time, what to skip, and why.

Claude Code ships with built-in MCP (Model Context Protocol) support. Add external tools — databases, APIs, SaaS platforms — directly from the terminal with claude mcp add. This page covers every command, config option, and common fix.

Here's the thing nobody tells you about Cursor's agent mode: it's brilliant at working with code and completely blind to everything your code talks to.

Last week we were debugging a webhook handler. Cursor had the code open, understood the control flow, spotted a race condition in the retry logic. Genuinely impressive. Then we needed to know whether the bug was actually hitting production — were customers seeing duplicate charges? The agent that just did 15 minutes of sophisticated code analysis couldn't answer a basic factual question about our own Stripe data.

So we opened a browser tab, logged into Stripe, searched for the customer, scrolled through PaymentIntents, compared timestamps manually, went back to Cursor, and typed what we found. The AI had all the context and none of the data.

We got tired of being the copy-paste bridge between our IDE and our dashboards.