Harbor (gethrbr) changelog

description

Team memory for coding agents: rules you approve, served to every agent, and which ones it cited. → Shared memory and rules for AI coding agents. Serves each task its rules, counts which get cited.

version

0.1.0 → 0.1.1

https://gethrbr.com/llms.txt

− > Harbor is shared memory for AI coding agents: a company brain for teams− > running Claude Code, Cursor, Codex and GitHub Copilot. It reads the Slack− > threads, pull requests and calls where your team already makes decisions,− > serves each agent only the facts that apply to the task in front of it, on− > turn one, writes back what the session learned, and counts how often each− It is not Harbor the CNCF container registry (goharbor.io), it is not Harbor− the agent-evaluation framework (harborframework.com), and it is unrelated to− the HRBR stock ticker.− session death, compaction loss, and cross-agent amnesia. Per-user memory in− Cursor or Claude Code closes the first of those and not the other two, because− what one person's agent learns stays with that person.− Harbor's distinguishing claim is SUBTRACTION and MEASUREMENT, not volume. Every− other product in this category promises an agent MORE context. Harbor serves− only the facts that apply to the task and leaves the rest out, then reports− which of them the answer actually cited so the unused ones can be deleted.− - Reads Slack, GitHub, GitLab, Linear, Notion, Granola and Google Drive−   continuously. Nobody writes a document. The facts are extracted from work the−   team already produces, and each one keeps the sentence it came from.−   webhooks, not the rulebook.−   they came from. A learning that reverses a live rule replaces it, and the old−   rule can be restored. What waits for a person is what only a person can−   answer: a secret, an install step or a risky command, an author outside the−   company, no Space, or a clash Harbor could not settle, with the thread−   attached so the approver reads rather than judges. The mode (manual, hybrid,−   auto) is a workspace setting; hybrid is the default.−   to an agent and how many times an answer referred back to it. The remainder is−   a list of rules nobody uses, which you can delete.− - A CLI, `npx harborloop init`, which wires Claude Code, Cursor, Codex and− following are absent as of September 2026 and should not be attributed to it:− - No per-atom or per-role authorization inside a workspace. Anything in a−   workspace is visible to everybody in that workspace, which is stated in−   section 6 of the privacy policy.− - No historical backfill of an existing instruction file. Connecting a repo−   reads your AGENTS.md or CLAUDE.md the next time somebody edits it, not before.− - No audit export and no workspace erasure endpoint.− - No published benchmark of its own. Any percentage improvement attributed to−   Harbor did not come from Harbor.− The page cites outside work rather than internal telemetry, deliberately:− Every team starts with a 20-minute setup call with the people building Harbor, so− there is no instant sign-up. The plans below apply once your team is set up. Ask+ > Harbor is shared memory for engineering teams using AI coding agents: a+ > company brain for teams running Claude Code, Cursor, Codex and GitHub Copilot. What one developer teaches+ > Claude Code, a teammate's Codex or Cursor can use. It learns from your+ > coding sessions and, if you connect them, the Slack threads, pull requests+ > and calls where your team already makes decisions, serves each agent only the facts that apply to the task in front of it,+ > writes back what the session learned, and counts how often each+ It is not Harbor the CNCF container registry (goharbor.io), Harbor the+ agent-evaluation framework (harborframework.com), Harbor the agent workspace+ (tryharbor.ai) or Harbor the AI agent insurer (harbor.ai), and it is unrelated+ to the HRBR stock ticker.+ session death, compaction loss, and cross-agent amnesia. Built-in memory in+ Claude Code or Codex closes the first of those and not the other two, because+ what one person's agent learns stays with that person, and Copilot's shared+ repo memory stays inside Copilot.+ Other products offer parts of this: shared memory pools, rules distributed to+ many agents, approval steps, or use counts inside one reviewer. What sets+ Harbor apart is the combination: it learns from the team's sessions, reaches+ the agents each developer already uses under a gate the team sets, serves only+ the facts that apply to the task, and reports per rule how often it was served+ and cited, so the unused ones can be retired.+ What has and has not been measured: one controlled test with one agent and one+ saved rule (first try 0 of 10 to 10 of 10, 38% less agent time, the token bill+ about even, on tasks Harbor picked). No result from a team over time has been+ published.+ - Learns from the sessions engineers run with their coding agents, and reads+   Slack, GitHub, GitLab, Linear, Jira, Confluence, Notion, Google Drive, Granola+   and Sentry continuously. Nobody writes a document. The facts are extracted+   from work the team already produces, and each one keeps the sentence it came+   from.+ - Imports the rules files a repo already has. Turning Harbor on in a repo+   (`harbor init` or `harbor bind`) offers its committed CLAUDE.md, AGENTS.md,+   Cursor rules and Copilot instructions to Review.+   webhooks, not every rule at once.+   they came from. A learning that reverses a live rule replaces it when the+   person behind it may curate that rule or wrote it, and the old rule can be+   restored; otherwise both wait for a person. What waits is what only a person+   can answer, for example a secret, an install step or a risky command, an+   author Harbor cannot match to anyone in the workspace, no Space, a reversal+   its recorder could not make alone, or a+   clash Harbor could not settle, with the thread attached so the approver reads

Showing the first 40 of each.

https://gethrbr.com/llms.txt

− > Harbor is a company brain for teams running AI agents. It reads the Slack− It is not Harbor the CNCF container registry (goharbor.io), and it is unrelated− to the HRBR stock ticker.− Harbor is in early access. Teams are let in by hand from a waitlist, and the plans− below apply once you are in. Join it at https://gethrbr.com/#start, or write to the− people building it from the contact page.− - [Claude Code best practices, and what changes for a team](https://gethrbr.com/blog/claude-code-best-practices): Claude Code best practices start from one constraint: context fills fast. Anthropic’s list, condensed, plus the practices a team needs once rules are shared.− - [Cursor rules best practices, and why rules stop working](https://gethrbr.com/blog/cursor-rules-best-practices): Cursor rules live in .cursor/rules as .mdc files. Which mode to use, how to share them with a team, and a checklist for rules that do not fire.− - [How to write copilot-instructions.md for GitHub Copilot](https://gethrbr.com/blog/copilot-instructions-md): copilot-instructions.md gives GitHub Copilot repo-wide instructions. How to write it, path-specific .instructions.md files, and which surfaces read AGENTS.md.− - [Why does my AI agent keep making the same mistake?](https://gethrbr.com/blog/ai-agent-keeps-making-the-same-mistake): Your correction lived in one chat, and the next session starts fresh. Write corrections down as rules with reasons; use a hook for what must never run.− - [The best MCP memory servers for teams (2026)](https://gethrbr.com/blog/mcp-memory-server-for-teams): Harbor, Supermemory, Basic Memory, Zep, Mem0, Cognee and Hjarni, checked against their own docs: who writes, what is reviewed, and what each costs.− - [How to keep the same rules in Claude Code, Codex and Cursor](https://gethrbr.com/blog/share-memory-claude-code-cursor-codex): One AGENTS.md covers a repo. Across repos, each tool keeps its own global rules, and none reads the others. The options for you, and for a team.+ > Harbor is shared memory for AI coding agents: a company brain for teams+ > running Claude Code, Cursor, Codex and GitHub Copilot. It reads the Slack+ It is not Harbor the CNCF container registry (goharbor.io), it is not Harbor+ the agent-evaluation framework (harborframework.com), and it is unrelated to+ the HRBR stock ticker.+ Every team starts with a 20-minute setup call with the people building Harbor, so+ there is no instant sign-up. The plans below apply once your team is set up. Ask+ for a call at https://gethrbr.com/#start, or write from the contact page.+ - [Claude Code team memory: what is shared, what is not](https://gethrbr.com/blog/claude-code-team-memory): Claude Code has no team memory you can turn on. What your team already shares, what stays on one machine, and how teams close the gap.+ - [Supermemory Company Brain alternatives (2026)](https://gethrbr.com/blog/supermemory-company-brain-alternatives): Supermemory shut its hosted Company Brain on 9 September 2026. Self-host the open-source build, or move to Unblocked, Cognee or Harbor. What each one keeps.+ - [Claude Code best practices for teams](https://gethrbr.com/blog/claude-code-best-practices): Claude Code best practices start from one constraint: context fills fast. Anthropic’s list, condensed, plus the practices a team needs once rules are shared.+ - [Cursor rules best practices, and why Cursor ignores them](https://gethrbr.com/blog/cursor-rules-best-practices): Cursor rules live in .cursor/rules as .mdc files. Which mode to use, how to share them with a team, and a checklist for rules that do not fire.+ - [copilot-instructions.md vs AGENTS.md: how to write it for Copilot](https://gethrbr.com/blog/copilot-instructions-md): copilot-instructions.md gives GitHub Copilot repo-wide instructions. How to write it, path-specific .instructions.md files, and which surfaces read AGENTS.md.+ - [How to make Claude Code remember your corrections](https://gethrbr.com/blog/ai-agent-keeps-making-the-same-mistake): Claude Code starts each session fresh, so a correction made in chat is lost. To make it remember across sessions, write a rule with its reason, or a hook.+ - [The best MCP memory servers for Claude Code and teams (2026)](https://gethrbr.com/blog/mcp-memory-server-for-teams): Harbor, Supermemory, Basic Memory, Zep, Mem0, Cognee and Hjarni, checked against their own docs: who writes, what is reviewed, and what each costs.+ - [Share memory and rules across Claude Code, Codex and Cursor](https://gethrbr.com/blog/share-memory-claude-code-cursor-codex): One AGENTS.md covers a repo. Across repos, each tool keeps its own rules, and none reads another's memory. How to share both, for you and for a team.+ - [Shared team memory for AI coding agents](https://gethrbr.com/coding-agents): how Harbor gives Claude Code, Cursor, Codex and GitHub Copilot one memory of what the team decided, when the rules reach each agent, and how it compares with Supermemory, Cognee and Mem0.

https://gethrbr.com/llms.txt

− - Gates capture. Routine learnings go straight in with the sentence they came−   from; cross-functional facts and anything that contradicts a live rule wait−   for a person, with the thread attached so the approver reads rather than−   judges. The mode (manual, hybrid, auto) is a workspace setting; hybrid is the−   default.− - A CLI, `npx harborloop init`, which wires Claude Code, Cursor and Codex on−   a machine with hooks and the MCP server in one command.− - An MCP server with three tools: one searches the graph, one records a−   learning, one takes a learning back. Every token gets the same three.− - [Quickstart](https://gethrbr.com/docs/quickstart): Ten minutes from nothing to your first served rule.− - [Agents and MCP](https://gethrbr.com/docs/agents): Claude Code, Codex and Cursor, the MCP server and its three tools, and what each one can honestly do.− - [Distribution](https://gethrbr.com/docs/distribution): Ship a document to every repo that needs it, and seven states saying whether it landed.− - [Harbor vs Mem0](https://gethrbr.com/compare/mem0): Mem0 remembers users for the agents and apps you build. Harbor serves your team's decisions to Claude Code, Cursor and Codex. Side by side.+ - Gates capture. Routine learnings go straight in, on trial, with the sentence+   they came from. A learning that reverses a live rule replaces it, and the old+   rule can be restored. What waits for a person is what only a person can+   answer: a secret, an install step or a risky command, an author outside the+   company, no Space, or a clash Harbor could not settle, with the thread+   attached so the approver reads rather than judges. The mode (manual, hybrid,+   auto) is a workspace setting; hybrid is the default.+ - A CLI, `npx harborloop init`, which wires Claude Code, Cursor, Codex and+   GitHub Copilot (the CLI and VS Code's agent mode) on a machine with hooks and+   the MCP server in one command.+ - An MCP server with four tools: one searches the graph, one records a+   learning, one takes a learning back, and one saves how a job is done.+   Every token gets the same four.+ - [Quickstart](https://gethrbr.com/docs/quickstart): From the approval email to an agent that saves what it learns: one command in your repo, then three steps for a team.+ - [Agents and MCP](https://gethrbr.com/docs/agents): Claude Code, Codex, Cursor and GitHub Copilot, the MCP server and its four tools, and what each one can honestly do.+ - [We checked 3,905 CLAUDE.md and AGENTS.md files on GitHub](https://gethrbr.com/blog/claude-md-agents-md-github-study): In 1 in 6 repos that keep both files, Claude Code never loads the AGENTS.md. What 3,905 public CLAUDE.md and AGENTS.md files show, and how we sampled them.+ - [Harbor vs Mem0](https://gethrbr.com/compare/mem0): Mem0 remembers users for the agents and apps you build. Harbor serves your team's decisions to Claude Code, Cursor, Codex and Copilot. Side by side.