The Lockstep blog
Keeping humans and AI coding agents in step: decision capture, context engineering, MCP, drift, and the record that files itself.
- 8 min read
ADRs for AI coding agents: how to make every agent read architecture decisions
Architecture decision records only help AI coding agents when the right record enters context at the right time. Use this setup for Cursor, Claude Code, and other agents.
Read → - 8 min read
Best MCP memory servers for teams using AI coding agents
Compare MCP memory servers for shared coding context, team knowledge, temporal facts, and approved decisions - then choose the smallest system that solves your actual memory failure.
Read → - 6 min read
CLAUDE.md vs agent memory: what each should store
CLAUDE.md, auto memory, task handoffs, and shared team decisions solve different context problems. Use this guide to decide what belongs where.
Read → - 5 min read
How to share memory between Cursor and Claude Code
Cursor and Claude Code keep different personal context. Use repo rules, task handoffs, live tools, and shared team decisions so work can move between them without a fresh explanation.
Read → - 5 min read
Cursor keeps losing context mid-project. How to stop re-explaining your codebase
If Cursor loses context after a long session or project switch, separate working context, stable rules, and team decisions so the next agent can continue without guessing.
Read → - 7 min read
Why Claude Code forgets architecture decisions - and how teams stop re-explaining them
Claude Code can understand a codebase and still lose the reasoning behind architecture decisions. Here is what to save, where to save it, and how to keep every agent aligned.
Read → - 7 min read
Who wrote this code? AI provenance starts with the wrong question
A commit can show who changed a file, and an attestation can show how an artifact was built. Neither explains why an AI agent made the change or which team decision it was meant to honor.
Read → - 7 min read
Two AI agents, one codebase: why parallel work breaks without shared memory
Two coding agents can each produce a reasonable change and still leave the product inconsistent. The missing layer is shared decision memory: the settled calls every agent must know before it acts.
Read → - 7 min read
Your AI coding agent forgets every decision overnight. Here's the fix.
AI coding agents start each session with a blank memory, so settled decisions get re-litigated at machine speed. Here's what that costs a team, and how persistent decision memory fixes it.
Read → - 9 min read
How we distill decisions from Slack noise
Inside Lockstep's distillation funnel: a four-test rubric, tiered models, prompt-cached extraction, and a human gate that nothing skips.
Read → - 9 min read
How cross-vendor agent capture works: hooks, MCP, and instruction files
Every major coding agent exposes the same three surfaces. How Lockstep uses lifecycle hooks, an MCP server, and instruction files to capture agent work.
Read → - 7 min read
The best tools for giving AI coding agents your team's context (2026)
Eight ways to give AI coding agents your team's context, ranked, from AGENTS.md to MCP servers to decision memory, with honest limits for each.
Read → - 7 min read
Best MCP servers for engineering teams (2026)
Nine MCP servers worth installing for a software team, ranked by usefulness: what each exposes, its access model, and where it falls short.
Read → - 7 min read
Best decision-log tools for product & engineering teams (2026)
Seven tools for keeping a decision log your team actually reads: Notion, Confluence, ADR files, Linear, GitHub Discussions, Loomio, and Lockstep.
Read → - 10 min read
The load-bearing hallway: what AI coding agents break first
Git, Jira, and PRDs were never the coordination layer — the human hallway was. AI coding agents delete it. What hospitals and remote teams teach us.
Read → - 7 min read
The engineering decision log: what to write down, and when
A practical guide to keeping a decision log: one question that tells you what counts, the four fields every entry needs, and the two moments teams always miss.
Read → - 15 min read
Architecture Decision Records: right idea, wrong reader
ADRs got the diagnosis right — record the decisions. But they quietly fail on the read side, and AI coding agents are about to make that failure loud.
Read → - 13 min read
AGENTS.md: what to put in it, and what it can't carry
A straight guide to what belongs in an AGENTS.md — commands, invariants, boundaries — what to keep out, and the coordination job no static file can do.
Read → - 5 min read
MCP vs RAG: When to Use Each
MCP and RAG both get outside information to a model, but they solve different problems. A plain comparison, a decision table, and when to use each.
Read → - 6 min read
The Feed Is Coming for GitHub
The friend graph, then the CRM, now GitHub: value keeps moving from the database up to the layer that reads across everything and decides what happens next.
Read → - 4 min read
Introducing Lockstep: A System of Record for AI Coding Agents
The tools that coordinate engineering work assume humans writing code. AI coding agents have no hallway conversations. Lockstep is the layer they're missing.
Read →