Field notes on running agents.
How we build, secure, and run a fleet of self-hosted AI agents.
One MCP Config for Claude Code, Cursor, and Codex
The short answer: one MCP config can be your source of truth, but it should not be one raw file copied into Claude Code, Cursor, and Codex. MCP standardizes…
AI Coding Agent Over-Editing and Diff Size Limits
AI coding agent over-editing is a reviewability failure, not a style annoyance. The agent touches more files than the task needs, bundles behavior with…
Coding Agent Action Bias and No Code Change Needed
The best coding-agent result is sometimes no diff. That sounds obvious until you measure it. FixedBench, a 2026 benchmark built from 200 already-fixed…
How to Prevent AI Coding Agents Modifying Tests
If an agent can rewrite the verifier, a green build stops meaning what your team thinks it means. The practical way to prevent AI coding agents modifying tests…
Claude Code LSP Setup for Production Repositories
The useful version of Claude Code LSP setup is not "turn on code intelligence and trust it." For a production repo, it is a small platform rollout: install the…
AGENTS.md vs CLAUDE.md vs Cursor rules for agent teams
<article> <p>A platform team adds Codex to one repo. A frontend team keeps using Cursor. A backend team standardizes on Claude Code. Within a…
Claude Code plan mode workflow for team approval gates
<article> <p>A senior engineer asks Claude Code to change a deployment workflow. The agent understands the request, finds the YAML file, edits a…
Playwright CLI vs MCP for Claude Code in production
<article> <header> <p>Claude Code can drive a browser in more than one way. That is useful, but it also creates a decision that teams should…
Claude Code checkpoints vs Git commits for team workflows
<article> <p>You ask Claude Code to clean up a tangled module. It touches five files, runs a formatter, tries a test command, changes direction…
Claude Code monorepo setup for large engineering teams
<article> <p>A monorepo team usually notices the problem in a small, annoying way. Claude Code starts confidently, reads a few files, then reaches…
AI coding agent database migration safety model
AI coding agent database migration safety starts with a simple question: what can the agent actually touch? For a platform or security engineer, the decision…
Claude Code skills vs subagents vs MCP for teams
Choosing between Claude Code skills vs subagents vs MCP is not a question of which feature is most powerful. For a senior engineer designing a team workflow,…
AI Coding Agent PR Reviewability Checklist for Teams
AI coding agents make pull requests cheaper to create. They do not make reviewer attention cheaper. For engineering managers and senior engineers, the…
Claude Code subagents best practices for team governance
Claude Code subagents best practices matter most when a team moves from individual experimentation to shared automation. A subagent can keep noisy research,…
Claude Code compaction strategy for long coding tasks
A good Claude Code compaction strategy treats compaction as a handoff boundary, not as durable memory. That distinction matters when a coding task spans hours,…
Self-hosted Coding Agent Runtime Comparison Guide
The right self-hosted coding agent runtime is the one that matches your trust boundary. For most security-conscious teams, that means per-task isolated…
Persistent Memory for AI Coding Agents: Governance
Persistent memory for AI coding agents should be treated as platform infrastructure, not as a bigger prompt. The useful design is layered: small repo…
MCP Context Bloat: How to Govern Tool Growth
MCP context bloat is the hidden tax of giving coding agents too many tools at once. Each enabled MCP server can add tool names, descriptions, input schemas,…
Claude Code Worktree Database Isolation Guide
Claude Code worktree database isolation means giving every agent its own runtime namespace as well as its own checkout. Git worktrees and Claude Code's…
AI Coding Agent Secrets: Protect .env and Tokens
AI coding agent secrets should not rely on .env files, ignore rules, or tool permissions as the main boundary. The workable pattern is layered containment:…
Multi-Agent Coding Orchestration Is Ops Now
Multi-agent coding orchestration works when agents run in parallel but review, isolation, approvals, and merge decisions stay controlled.
Coding Agent Dependency Security: Privileged Installs
Coding agent dependency security starts by treating package installs as privileged operations, with policy approval, lockfiles, and registry controls.
AGENTS.md Best Practices for Coding Agents
AGENTS.md best practices for coding agents: what to include, what to leave out, how to handle monorepos, and where enforcement belongs.
Claude Code Hooks for Permissions: A Safer Pattern
Claude Code hooks for permissions can reduce approval fatigue when paired with narrow allowlists, sandboxing, managed settings, and audit logs.
MCP Server Monitoring: What Reliability Requires
MCP server monitoring needs more than uptime checks. Track tool discovery, auth, latency, retries, timeouts, and per-tool safety controls.
Cloud Coding Agent vs Local CLI Agent: Runtime Guide
Cloud coding agent vs local CLI agent is really a runtime decision: compare ownership, security, secrets, persistence, cost, and PR handoff before rollout.
Coding Agent Evaluation Metrics for Real Repos
Coding agent evaluation metrics should measure repo tasks, review quality, cost, regressions, partial progress, and oversight beyond leaderboard resolve rate.
Coding Agent Cost Monitoring: Control Spend Early
Coding agent cost monitoring helps platform teams control token, credit, and workflow spend by combining budgets, telemetry, attribution, and policy.
Securing Coding Agents in CI/CD: Practical Baseline
Securing coding agents in CI/CD starts with one rule: treat untrusted GitHub content as hostile, then limit secrets, tokens, network, and write access.
AI Code Review Agent: Evaluation Guide for PRs
An AI code review agent should be evaluated as a governed PR workflow, not a comment bot. Compare signal, cost, permissions, context, and human review impact.
Credential Proxy for AI Agents Without Secret Exposure
A credential proxy for AI agents lets agents call private repos, APIs, CLIs, and MCP tools without exposing raw secrets to runtimes, prompts, or logs.
Agent Harness for Coding Agents: Runtime Architecture
An agent harness for coding agents controls sandboxes, state, permissions, tool execution, review flow, lifecycle, and cleanup around safe AI coding work.
AI agent observability for safe coding agent rollouts
AI agent observability for coding agents: trace runs, commands, diffs, tests, costs, approvals, and risky side effects before code reaches production.
Long-running background AI agents need durable workers
Long-running background AI agents need durable workers, queues, checkpoints, approvals, sandboxing, cost caps, observability, PR review, and control.
Persistent AI Agent Workspace Architecture Guide
Design a persistent AI agent workspace with durable files, sandbox snapshots, memory boundaries, rollback, tenant isolation, and clear retention policy.
Self-hosted coding agent runtime: build, buy, operate
A self-hosted coding agent runtime gives policy control and data residency, but shifts sandboxing, secrets, audit, cleanup, and capacity to your team.
Production Workflows for AI Coding Agents That Scale
Production workflows for AI coding agents need isolated workspaces, reviewable diffs, CI controls, protected merge gates, and accountable human ownership.
MCP vs function calling: Practical architecture guide
MCP vs function calling explained for engineers choosing between direct tool calls, MCP servers, runtime discovery, auth boundaries, latency, and reuse.
MCP Security for AI Agents: Production Controls
MCP security for AI agents needs token audience checks, sandboxed tools, schema pinning, approval UX, egress limits, and clear incident-ready audit trails.
AI Agent Sandboxing: Secure Coding Agent Controls
AI agent sandboxing helps security leaders control source access, secrets, network egress, and coding agent execution risk before agents touch private code.