BLOG

Thoughts on AI infrastructure

Engineering deep-dives, security breakdowns, and practical guides for building production AI systems.

How to query your database in plain English (without writing SQL)

Your database has the answers. Your team doesn't have time to write queries. Here's how MCP closes that gap in 4 minutes.

Read
AI7 min read

Semantic layers for AI database agents: stop making the model infer business meaning from table names

AI database agents need a semantic layer for metrics, entities, joins, freshness, and approved definitions. Table names alone are not enough for trustworthy answers.

May 18, 2026
Read
Security7 min read

Result redaction for AI database agents: hide sensitive fields before the model summarizes them

AI database agents should not receive every field a query can return. Result redaction keeps sensitive columns, samples, and identifiers out of model context by default.

May 18, 2026
Read
Security7 min read

Row-level security for AI database agents: make data boundaries impossible to forget

AI database agents should not rely on remembered tenant filters. Row-level security, approved views, and scoped roles make data boundaries enforceable below the model.

May 17, 2026
Read
MCP7 min read

Query routing for AI database agents: send the question to the right source before SQL runs

AI database agents need query routing before execution. Some questions belong on live databases, some on replicas, some on warehouses, and some should fail closed.

May 17, 2026
Read
Governance7 min read

Dry-run mode for AI database agents: preview the blast radius before anything changes

AI database agents need dry-run workflows before writes, exports, and broad queries. A safe preview shows affected rows, policy checks, and rollback context before execution.

May 16, 2026
Read
AI7 min read

Freshness windows for AI database answers: stop treating stale context as live truth

AI database answers need freshness windows. Production teams should show when data was read, which snapshot was used, and when stale context must fail closed.

May 16, 2026
Read
Security7 min read

AI database query budgets: the production guardrail most natural-language SQL demos skip

AI database agents need query budgets for rows, time, cost, scope, and retries. Without budgets, natural-language SQL can become an unbounded production risk.

May 15, 2026
Read
Governance7 min read

Approval gates for AI database writes: keep the model out of the final decision loop

AI agents should not approve their own database writes. Production MCP workflows need external approval gates, scoped tools, and audit trails for side effects.

May 15, 2026
Read
Security7 min read

Fail-closed MCP database tools: how AI agents should handle unsafe or unclear queries

MCP database tools should fail closed when scope, permissions, freshness, or query intent is unclear. Helpful failure modes are part of production AI safety.

May 14, 2026
Read
AI7 min read

Metric definitions for AI database agents: stop making the model guess what revenue means

Natural-language SQL is only useful when the agent knows your business metrics. Table names are not enough for trustworthy AI reporting.

May 14, 2026
Read
MCP8 min read

Tool result contracts for AI database agents: make answers debuggable before they are summarized

AI database agents need structured result contracts, not just raw rows, so teams can debug wrong answers, enforce limits, and trust natural-language reporting.

May 13, 2026
Read
Security7 min read

Tenant scoping for AI database agents: the filter that cannot be optional

AI database agents can answer useful business questions, but multi-tenant data access needs enforced tenant scoping before natural-language SQL reaches production.

May 13, 2026
Read
PostgreSQL8 min read

MCP server for PostgreSQL: a production checklist before you connect AI

Before connecting Claude, ChatGPT, or other AI clients to PostgreSQL through MCP, teams should define scopes, read-only access, query limits, context, and audit trails.

May 12, 2026
Read
AI7 min read

Schema context for AI database agents: what the model needs before it queries

AI database agents need more than a connection string. Good schema context turns natural-language questions into safer, narrower, more useful database queries.

May 12, 2026
Read
MCP6 min read

MCP Tool Search for database agents: discovery is not permission design

MCP Tool Search can reduce context bloat, but database-connected agents still need narrow tools, explicit permissions, and audit trails before discovery reaches production.

May 11, 2026
Read
MCP7 min read

MCP schema drift: why database agents need stable tool contracts

When an MCP tool schema changes, the agent's behavior can change too. Database-connected agents need contract review, schema context, and runtime controls before drift reaches production.

May 10, 2026
Read
Security6 min read

Audit-ready MCP database workflows: what evidence to capture

AI database access should leave a reviewable trail. Here is what teams should capture when MCP tools answer questions from live production data.

May 10, 2026
Read
Security6 min read

Data minimization for AI database agents: return less by default

AI agents do not need unlimited rows to be useful. Data minimization, approved views, limits, and redaction should be part of every production MCP database setup.

May 9, 2026
Read
AI6 min read

Agent memory for database workflows: useful context or hidden risk?

Long-term agent memory can improve database workflows, but teams need rules for what is stored, retrieved, redacted, and audited.

May 8, 2026
Read
Security6 min read

Short-lived credentials for AI database agents: reduce the blast radius first

AI agents should not hold broad, long-lived database credentials. Use short-lived, scoped access with tool boundaries, query limits, and audit logs.

May 8, 2026
Read
Security7 min read

Read-only AI analytics: why SELECT-only is necessary but not enough

Read-only access is the right default for AI analytics, but production teams still need scope, schema context, result limits, and audit logs.

May 7, 2026
Read
Security7 min read

AI database connector architecture: the five boundaries teams should define first

Connecting AI to a database is easy to demo. Production teams need five boundaries before Claude or ChatGPT can safely answer live data questions.

May 7, 2026
Read
Use case7 min read

Repeatable AI reporting workflows: when one-off database questions are not enough

One-off AI database answers are useful. The bigger operational win comes when teams turn recurring questions into repeatable MCP-powered reporting workflows.

May 6, 2026
Read
Security7 min read

Azure SQL MCP server: how to give AI agents useful access without broad cloud permissions

Azure SQL often holds the operational answers teams need. The safe path is not broad cloud access — it is scoped MCP tools, read-only roles, and auditable queries.

May 6, 2026
Read
Security6 min read

MCP tool descriptions are a security boundary, not documentation garnish

For AI agents, tool descriptions shape behavior. In production MCP servers, naming, schema design, and constraints become part of the safety model.

May 5, 2026
Read
Comparison7 min read

PostgreSQL MCP alternatives: build, open source, or managed infrastructure?

Teams connecting AI agents to PostgreSQL usually compare three paths: build a custom MCP server, run open-source tooling, or use managed MCP infrastructure.

May 5, 2026
Read
Comparison7 min read

ChatGPT database connector alternatives: MCP, SQL chatbots, and custom APIs compared

Teams want ChatGPT to answer questions from live data. The real decision is whether to use a SQL chatbot, a custom API, or an MCP database connector.

May 4, 2026
Read
Comparison7 min read

MCP vs REST API for AI agents: why tools beat endpoints for live data

REST APIs were designed for applications with predictable flows. AI agents need a tool layer that carries intent, scope, context, and auditability.

May 4, 2026
Read
Tutorial7 min read

MCP server for PostgreSQL: how AI agents can query live data safely

PostgreSQL already holds the answers many teams need. An MCP server gives AI agents a controlled way to ask for them without building another custom backend.

May 3, 2026
Read
Security7 min read

Secure AI database access: the checklist before you connect production data

AI database access becomes useful fast. It also becomes risky fast unless teams define scope, permissions, schema context, and auditability before rollout.

May 3, 2026
Read
Tutorial7 min read

Claude MCP database setup: from weekend prototype to production rollout

Connecting Claude to a database is easy to demo. The real work is turning that demo into a controlled, repeatable production setup.

May 2, 2026
Read
Industry POV7 min read

AI SQL assistant vs MCP database server: the architecture difference teams miss

An AI SQL assistant can help write queries. An MCP database server gives AI tools a controlled way to use live data. Those are not the same thing.

May 2, 2026
Read
Use case7 min read

Internal reporting with AI and MCP: fewer data tickets, better weekly answers

Most internal reporting requests are not complex. They are recurring, contextual, and slow because the data sits behind SQL, APIs, and team boundaries.

May 1, 2026
Read
Tutorial7 min read

MCP server for SQL Server: how to give AI agents safe access to enterprise data

SQL Server still runs critical business data. Here is how an MCP server can make that data useful to AI agents without turning production into an experiment.

May 1, 2026
Read
Product7 min read

Azure MCP tools for AI agents: expose cloud operations without exposing everything

Azure environments are full of useful operational context. The challenge is giving AI agents the right Azure tools through MCP without turning every server into an all-access cloud console.

Apr 30, 2026
Read
Product7 min read

Scheduled MCP Flows: turning AI database answers into repeatable reporting workflows

One-off AI database questions are useful. Scheduled MCP Flows are how teams turn those questions into repeatable reports, checks, and operational routines.

Apr 30, 2026
Read
Security6 min read

AI database access governance: what to decide before you connect Claude to production

AI database access needs governance before it needs enthusiasm. Decide scope, roles, logging, and ownership first — then connect your MCP clients to live data.

Apr 29, 2026
Read
Tutorial7 min read

MCP server for MySQL: how to let AI query live data without building a custom backend

MySQL already holds the answers your team asks for every week. An MCP server gives Claude, ChatGPT, and other AI clients a governed way to query it without another pile of custom endpoints.

Apr 29, 2026
Read
Comparison7 min read

Custom API vs MCP for AI agents: when building another endpoint is the wrong move

A custom API can expose data to an app. AI agents need something more discoverable: tools, schema context, guardrails, and auditability. That is where MCP changes the architecture.

Apr 28, 2026
Read
Security6 min read

Scoped database access for AI agents: the guardrail most teams skip

AI agents should not get a master key to production data. Scoped database access gives them enough context to answer questions without turning every prompt into a security review.

Apr 28, 2026
Read
Comparison7 min read

MCP vs SQL chatbot: why the protocol matters more than the chat box

A SQL chatbot can translate text into queries. MCP gives AI agents a governed way to discover tools, understand schemas, and use database access safely.

Apr 27, 2026
Read
Security6 min read

Audit AI database queries before they become a compliance problem

AI database access is only safe if every query can be traced. Here is what audit logging needs to capture when teams connect MCP clients to production data.

Apr 27, 2026
Read
Engineering6 min read

Natural language SQL fails when the AI cannot see your schema

The hard part of AI database querying is not translating English into SQL. It is knowing what your tables mean. Schema context is what turns a clever demo into a reliable workflow.

Apr 26, 2026
Read
Industry POV6 min read

Your REST API was built for apps. Your AI agent needs tools.

REST APIs are excellent for software. AI agents need something more contextual: discoverable tools, clear schemas, and scoped actions. MCP is the layer that turns APIs into usable AI infrastructure.

Apr 26, 2026
Read
Tutorial7 min read

How to connect ChatGPT to PostgreSQL without building a custom API

Most teams do not need another internal API just so an AI assistant can answer database questions. MCP gives you a cleaner path from PostgreSQL to ChatGPT.

Apr 25, 2026
Read
Security6 min read

MCP read-only database access: how to give AI answers without giving it production risk

Connecting AI to a live database sounds risky. It is — unless the MCP layer is designed around read-only access, scoped tools, and auditability from day one.

Apr 25, 2026
Read
Use Case4 min read

Use Case: Fleet Reporting & Natural Language SQL with MCP

Fleet teams should not wait on analysts just to answer operational questions. Here's how MCP makes live fleet reporting available in plain English for non-technical staff.

Apr 13, 2026
Read
Industry POV4 min read

Why AI projects stall at the database layer

Most AI projects do not fail because the model is bad. They stall because every useful answer still depends on manual SQL, schema checks, and data-team handoffs.

Apr 7, 2026
Read
Tutorial7 min read

From zero to AI-powered database queries in 10 minutes

A step-by-step tutorial for connecting Claude (or any MCP-compatible AI) to your PostgreSQL or MySQL database using Conexor — no custom code required.

Apr 4, 2026
Read
Security4

The patch you can't apply to the device you don't know exists

Your security tools are only as effective as the inventory they're working from. If your visibility is incomplete, your protection is incomplete.

Apr 3, 2026
Read
Use Case5 min read

Kill the data request ticket: how AI query layers are changing engineering workflows

Your data team spends 40% of their week on requests that should take seconds. Here's how MCP-based AI query layers are eliminating the bottleneck — and what it means for your team.

Apr 2, 2026
Read
Tutorial6 min read

How to connect MySQL to Claude (and ask it anything in plain English)

MySQL has your data. Claude has the intelligence. The missing piece is MCP — and it takes about 5 minutes to set up. Here's exactly how.

Apr 1, 2026
Read
Comparison3 min read

Elementor MCP vs conexor.io: website builder AI vs. database AI

Elementor's AI works on your WordPress site. Conexor's MCP connects your database to Claude. If you're searching for a way to query your data with AI — here's the right tool.

Apr 1, 2026
Read
Comparison4 min read

Windsor MCP vs conexor.io: MCP for marketing data vs. your own database

Windsor.ai connects marketing platforms to AI via MCP — ad spend, attribution, campaign data. Conexor connects your own databases. Different data, different use cases.

Apr 1, 2026
Read
Comparison4 min read

Komodor MCP vs conexor.io: which one actually connects your data to AI?

Komodor's MCP server is great for Kubernetes ops. But if you need your databases — PostgreSQL, MySQL, SQL Server — talking to Claude or Cursor, that's a different tool for a different job.

Apr 1, 2026
Read
MCP5 min read

Why your AI assistant can't answer business questions (and how to fix it in 5 minutes)

You have Claude. You have GPT-4. You have Cursor. But when someone asks "what's our churn this month?" — your AI goes blank. Here's why, and how to fix it.

Apr 1, 2026
Read
Explainer6 min read

What is MCP (Model Context Protocol) and why does it matter for your business?

MCP is Anthropic's open protocol for connecting AI assistants to external data and tools. Here's what it means for businesses that want AI to actually use their data.

Mar 10, 2026
Read
Use Case5 min read

Replacing weekly reports with AI: how operations teams use conexor.io

Operations managers used to wait until Monday for last week's numbers. Here's how teams use conexor.io to get any metric, on demand, in plain English.

Mar 7, 2026
Read
Security8 min read

How conexor.io enforces zero data exposure — even if our servers are compromised

A deep dive into our credential encryption architecture. TL;DR: your connection strings are AES-256 encrypted with a key we never store next to the data, so a breach of our control plane reveals nothing usable.

Feb 12, 2026
Read
Engineering6 min read

Why we moved from string substitution to parameterized queries — and why it matters for AI

AI models generate SQL. That's a prompt injection attack waiting to happen. Here's how our protocol-level parameterization makes SQL injection structurally impossible, regardless of what the model generates.

Jan 28, 2026
Read
Tutorial10 min read

MCP explained: what it is, why it matters, and how to use it with your database

Model Context Protocol is the missing layer between AI models and enterprise data. We explain what it actually is (not the marketing version), how it works under the hood, and why it's the right abstraction.

Jan 15, 2026
Read
Engineering7 min read

Inside schema discovery: how we turn 500 tables into useful AI tools in under 30 seconds

Auto-generating MCP tools from a production database isn't magic — it's careful introspection, batching, and type-mapping. Here's how the sausage is made, and what we do to avoid tool overload.

Dec 20, 2025
Read
Architecture5 min read

On-premise agent vs. cloud connector: which should you use?

Most teams don't need to run the agent on-prem. But if your security team requires it, here's exactly what changes — what data leaves your network, what stays, and what the latency trade-offs are.

Dec 5, 2025
Read
Compliance9 min read

Building a compliance-ready AI stack: what your audit log actually needs to contain

SOC 2, HIPAA, and GDPR all have different requirements for AI-generated queries. Here's what your audit trail actually needs to contain to satisfy all three frameworks.

Nov 18, 2025
Read
Relay

Quick questions

Relay

Quick questions

Ask me