Blog

Insights from the Glass Box

Field notes on agentic AI, the Model Context Protocol, observability, and building transparent, portable AI workflows.

Integrations

Give your agent GitHub access, not your keys

A repo-scoped token can force-push to main. The safer way to connect GitHub to an agent: scoped verbs, a server-side key, and a receipt for every call.

Read article →
Agent infra

It worked in Cursor. Then you shipped it.

The demo worked because you were the safety system. Production takes you out of the loop — here's the structural layer that has to replace you.

Read article →
Engineering

We let an AI write to production. Then a retry wrote twice.

We took the human out of the deploy loop, and a dropped network reply published the same post twice — plus the two checks that keep an agent's writes to production safe.

Read article →
Reliability

The agent runs while you sleep. Design for that.

No one's watching a 3 a.m. run. What keeps a scheduled agent safe: enforcement outside the model, approvals, idempotent writes, and a receipt by morning.

Read article →
Integrations

The right way to hand your AI agent a spreadsheet

Most ways to wire a sheet to an agent are rented glue that breaks. Give it a few exact verbs you own — key server-side, a policy on every write.

Read article →
Security

Your AI agent's guardrails shouldn't live in the prompt

Prompt rules are advisory — a model can ignore them. The guardrail that holds runs server-side, in front of the action, with a receipt for every attempt.

Read article →
Engineering

We deleted the session. The agents got more reliable.

An "active app" the server remembers is lovely in a demo and a bug in production. Why we made every tool call carry its own scope.

Read article →
Governance

An audit trail your AI agent can't skip

Logging is what your code writes when it remembers. An audit trail is a record the agent can't avoid producing — and where it's made decides if you can trust it.

Read article →
How-to

Your AI helpdesk should close the ticket, not draft a reply

Triage and deflection are the easy half. The valuable half is an agent that safely takes the action, resolves the ticket, and logs every step.

Read article →
Insights

Why AI agents break in production

The demo runs one happy user on clean data. Production adds tenancy, rate limits, drift, and cost — and the safe write path that closes the gap.

Read article →
Engineering

The page you're reading is a database record

How OBTO hosts its own website on OBTO: pages as rows, line-level patches, sha256-gated deploys, and the one build step we kept on purpose.

Read article →
How-to

Give your AI agent a seat in Slack

What "connect Slack" really means, the two paths to get there, the scopes that matter, and the guardrails that let an agent post without keeping you up at night.

Read article →
Guides

Self-hosted AI agent platforms: a buyer's guide

What "self-hosted" really means, a fair tour of the open-source and managed options, and the tradeoffs that decide which platform you own.

Read article →
Security

Your AI agent should never hold its own keys

Every useful agent is a small pile of secrets. Where API keys leak — prompt, client, logs — and the one pattern that keeps them server-side.

Read article →
Engineering

The Harness and the Substrate

How we built an agent runtime on a data-native platform: wrap-and-tee observation off the critical path, app-as-data, and what we've proven vs designed for.

Read article →
Guides

Connect Postgres to an AI agent without handing it raw SQL

The safe pattern for database access: expose scoped, typed tools instead of raw SQL, keep credentials server-side, and log every query.

Read article →
Engineering

How we built the Glass Receipt

Every AI run should leave a receipt. Why we gave each run an itemized, queryable ledger of models, tokens, cost, and tools, on by default.

Read article →
Engineering

Know your AI cost per run, before the bill

Per-token pricing hides the bill until it's too late. Route easy calls to a cheap model, keep the strong one for hard calls, and see each run's cost as it happens.

Read article →
Guide

What is an MCP server?

The plain-English guide: what an MCP server is, the three things it exposes, and how local and remote servers differ.

Read article →
Engineering

A prototype that feels like a product still isn't one

A demo proves the idea. Production is the dozen layers underneath it — and why AI codegen makes the gap bigger, not smaller.

Read article →
Vision

Why we built OBTO headless-first

Most AI platforms start with a chat box. We started with the tools and the audit trail, and treated the interface as optional — here's why.

Read article →
Integrations

Connect ServiceNow to AI Agents with MCP

Wrap the ServiceNow Table API as narrow, least-privilege MCP tools an agent can call safely: credentials stay server-side, and every call lands on the Glass Receipt.

Read article →
Engineering

The Big AI Lock-In — and How We Engineered Around It

Models get deprecated, repriced, and knocked offline. Here's how we keep memory and the live workstream outside any single model or client — so the work stays yours.

Read article →
Engineering

MCP Server as a Service: Your Data, Behind Your Firewall

Every AI client now speaks MCP. The question is whose server exposes your business data — and whether it runs behind your firewall, governed and battle-tested.

Read article →
Engineering

OBTO vs Lovable & Base44: Vibe Coding You Can Actually Own

Where the AI app builders shine, where they keep your software inside their walls, and what "Describe it. Ship it. Own it." actually means.

Read article →
Engineering

OBTO vs LangSmith vs Langfuse: Choosing a Self-Hosted AI Stack

Where the three tools actually overlap, what self-hosting really costs for each, and an honest guide to choosing your agent stack.

Read article →
Engineering

LLM Data Cleaning: Patterns for Production ETL Pipelines

Rules first, model second, humans last — hybrid cleaning pipelines, entity resolution without O(n²) costs, validation gates, and cost control.

Read article →
Engineering

AI Ticket Triage: How Agents Automate the Helpdesk

Where triage agents actually work, the guardrails they need, and the three metrics that prove helpdesk automation is paying off.

Read article →
Engineering

AI Agent Cost Tracking: Know What Every Run Costs

Why agent loops make token math non-linear, how to compute cost per completed task, and the five cost levers ranked by impact.

Read article →
Engineering

Multi-Model Orchestration: Routing LLMs by Cost and Task

Route each request by cost, latency, and task instead of paying frontier prices for everything — the strategies that work, and the hidden costs.

Read article →
Engineering

How to Host a Remote MCP Server in 2026: A Practical Guide

Transport, OAuth 2.1, scaling, and observability — what it actually takes to keep a remote MCP server alive in production.

Read article →
Engineering

How to Build an MCP Tool: A Practical Guide

The three MCP primitives, stdio vs. Streamable HTTP transport, a worked example, and what it takes to run a tool in production.

Read article →
Engineering

AI Agent Observability: A Practical Guide for Teams

What observability actually requires for agents — traces, tool calls, per-step token cost, and replay — and how to instrument it.

Read article →
Vision

How OBTO is Defining the AI Workforce of the Future

How transparent, observable agents are reshaping the way teams build and deploy an AI workforce.

Read article →
Engineering

Support for Groq: Fast Inferencing for Better AI Workflows

Bringing Groq's high-speed inference to OBTO to make agentic workflows faster and more cost-efficient.

Read article →
Architecture

Design & Architecture

A visual walkthrough of the OBTO Glass Box architecture — how the runtime, observability, and MCP layers fit together.

View board ↓
Architecture

Design & Architecture