--- title: "Cork vs Claude Code agent teams: one session's crew, or every agent on the repo · Cork" description: "Cork vs Claude Code subagents, agent teams and cross-session messaging: what each coordinates, where agents collide, and when to use both together." canonical: "https://corkcli.com/blog/cork-vs-claude-code-agent-teams" last_updated: "2026-11-14" published: "2026-10-03" author: "Kevin Vera" --- # Cork vs Claude Code agent teams: one session's crew, or every agent on the repo Cork vs Claude Code subagents, agent teams and cross-session messaging: what each coordinates, where agents collide, and when to use both together. By Kevin Vera · October 3, 2026 · [Comparisons](https://corkcli.com/blog/c/comparisons) · 8 min read How is Cork different from Claude Code's built-in subagents, agent teams and cross-session messaging? Claude Code's built-in tools let one person's Claude sessions split work, run side tasks and pass notes to each other. Cork is a [coordination layer](https://corkcli.com/blog/coordination-layer-ai-coding-agents) underneath every agent on a repository, including Codex, Cursor and teammates' machines, that briefs each agent on the code it is about to touch and holds an edit that would land on code another agent is changing right now. Put simply: Claude Code organizes your own crew; Cork keeps every crew on the codebase from overwriting each other, and replaces none of Claude Code's features. ## Cork vs Claude Code agent teams at a glance | Topic | Cork | Claude Code subagents, agent teams and messaging (as of October 2026) | |---|---|---| | What it is | Coordination layer for many coding agents and the people running them on one codebase | Built-in ways for Claude Code to run several Claude sessions: subagents, agent teams (experimental), cross-session messaging | | Where agents run | Wherever they already run (your machine); Cork does not run agents | On your machine, or in the cloud for cloud sessions | | Agents covered | Claude Code, Codex, Cursor, plus any MCP client | Claude sessions only ("In every approach the workers are Claude sessions") | | Who is covered | You, and with Team, your teammates' machines | Your own sessions; messaging reaches your sessions on other machines via Remote Control | | What is shared | Signatures, callers and recent changes of code being touched, live change notices, durable notes anchored to code | Plain-text messages, a shared task list (agent teams), subagent summaries | | Same-file conflict handling | Write gate holds an edit against code another agent is changing; [semantic lint](https://corkcli.com/blog/semantic-lint) after renames | Docs advise partitioning work or using worktrees: "Two teammates editing the same file leads to overwrites" | | Install | `curl -fsSL https://corkcli.com/install.sh \| sh` | Built in; agent teams need `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` | | Pricing | Free (local, and 10 agent hours a month in the cloud); $20 a month per 100 agent hours, pooled | Included with paid Claude plans; Pro from $17/month billed annually | ## What is Claude Code's built-in multi-agent tooling? Claude Code's docs list five ways to run work in parallel (subagents, agent view, agent teams, dynamic workflows and projects) plus supporting tools like worktrees and cross-session messaging. Three matter most here. - **Subagents** are "specialized AI assistants that handle specific types of tasks." Each runs in its own context window and returns a summary to the caller. A subagent can run in its own git worktree with `isolation: worktree`. - **Agent teams** coordinate multiple Claude Code instances: a lead session, teammates in their own context windows, a shared task list with dependencies, and a mailbox so teammates message each other directly. They are "experimental and disabled by default," and the docs note they "use significantly more tokens than a single session." - **Cross-session messaging** lets Claude "list and message your other Claude Code sessions on this machine," and reach sessions on other machines or in the cloud through Remote Control. The docs give the exact use case Cork cares about: "When a change in one session breaks what another is building on, Claude can warn that session before you notice." Messages are plain text, and Claude decides when to send them. On same-file edits the docs are direct. For agent teams: "Two teammates editing the same file leads to overwrites. Break the work so each teammate owns a different set of files." Agent teams do not isolate teammates in worktrees. ## What is Cork? Cork (corkcli.com, by Lithometric) keeps several AI coding agents, and the people running them, from overwriting each other on one codebase, and hands each agent what the others already learned. It does not run agents or host machines. Hooks in Claude Code, Codex and Cursor run the protocol with no extra model calls, and an MCP server at https://mcp.corkcli.com works with any MCP client. - **Briefing:** when an agent starts or gets a prompt, it receives the exact signatures, callers and recent changes of the code it is about to touch. - **Live awareness:** when another agent renames or re-signatures code yours depends on, your agent hears it on its next step. - **[Write gate](https://corkcli.com/blog/write-gate):** an edit against code another agent is changing right now is held until your agent has their version. - **Semantic lint:** a teammate's rename or signature change becomes a lint rule your agents are checked against. - **Landing:** `git push` becomes one step that rebases onto teammates' commits, reruns the tests and pushes. - **Agent messaging and MCP tools:** message one agent, all of them, or whoever is working on a function; get_symbol, blast_radius, recap, remember/recall. ## Where Claude Code's built-in tools stand out - **Zero setup for subagents and messaging.** They ship with Claude Code. Cross-session messaging is on with nothing to enable on supported versions. - **Real orchestration.** Agent teams give you a lead, a task list with dependencies, self-claiming with file locking, and hooks like `TaskCompleted` for quality gates. Cork does not plan or assign work. - **Context hygiene.** Subagents keep noisy searches and logs out of your main conversation and return only a summary. - **Cloud and cross-machine reach.** With Remote Control, messages reach your cloud sessions and your other machines. ## Where Cork wins - **Not just Claude.** Codex and Cursor agents on the same repo are covered by the same protocol. Claude Code's tools coordinate Claude sessions only. - **Not just you.** On the paid plan, agents on teammates' machines see each other. Cross-session messaging is scoped to your own sessions. - **Enforcement, not etiquette.** Claude Code relies on Claude choosing to send a warning, or on you partitioning files. Cork's write gate holds the conflicting edit, and semantic lint checks agents against a teammate's rename. - **Context without asking.** Every agent gets a briefing on the code it is about to touch, computed by hooks with no extra model calls, rather than relying on another session to remember to send a note. - **Merge time.** Worktree-isolated subagents still meet at merge. Cork's landing step rebases onto teammates' commits and reruns tests before pushing. ## Can you use both? Yes, and this is the most common setup. Keep using subagents for side tasks, agent teams for planned parallel work, and cross-session messaging for ad hoc notes. Cork's hooks run inside each Claude Code session, so every subagent, teammate and solo session on the repo gets briefings and passes through the write gate, alongside any Codex or Cursor agents and your teammates' agents. ## Pricing compared Claude Code is included in every paid Claude plan as of October 2026: Pro at $17/month billed annually ($20 monthly), Max from $100/month, Team at $20/seat/month for Standard seats, and Enterprise. Agent teams use more tokens because each teammate is a separate Claude instance. Cork Free runs locally with unlimited agents and repos on one machine and no account; Cloud Free gives each account 10 agent hours a month (30 bonus hours the first month) and a 30-day history window. Beyond that you pay for agent hours, not seats: $20 a month per unit of 100 hours, pooled across the workspace, graduated to $17 (units 11 to 50) and $15 (51 and up), with 15% off a year. Enterprise starts from about 100 units. In Cork's benchmark (12 Claude Code agents at once on one repository, hard tasks), agents with Cork used 53% fewer tokens than the same agents without it. ## Bottom line Claude Code's subagents, agent teams and messaging are excellent for one person orchestrating Claude sessions, and you should use them. They leave same-file safety to partitioning and to Claude remembering to send a message. Cork adds a protocol underneath that covers every agent and every teammate on the repo, enforces it at write time, and works whatever tool launched the agent. ## Frequently asked questions ### Is Cork an alternative to Claude Code agent teams? Not really. Agent teams plan and assign work among Claude sessions. Cork does not plan or run agents; it coordinates whatever agents are already running so they do not overwrite each other. Most people use both. ### Do Claude Code agent teams stop agents from overwriting each other? Not by themselves. The docs say "Two teammates editing the same file leads to overwrites" and recommend giving each teammate a different set of files. Cork's write gate holds an edit against code another agent is changing right now. ### Can multiple Claude Code agents talk to each other? Yes. Teammates message each other directly, named subagents can be messaged with `SendMessage`, and cross-session messaging lets your separate sessions exchange plain-text messages. Cork adds messaging across tools and teammates, including to whoever is working on a given function. ### Does cross-session messaging work with Codex or Cursor? Claude Code's docs say that in every approach "the workers are Claude sessions," and other tools are reached through MCP. Cork's hooks cover Claude Code, Codex and Cursor, and its MCP server works with any MCP client. ### Do I still need worktrees with Cork? Worktrees isolate files while agents work, but the work still meets at merge time. In a shared checkout Cork gates conflicting writes; across worktrees it tells agents about teammates' changes and makes landing a rebase plus test run. ## Sources Checked October 2026: - - - - - - - ## Related reading - [How to run multiple Claude Code agents on one repo without conflicts](https://corkcli.com/blog/run-multiple-claude-code-agents-one-repo): Run several Claude Code agents on one repository safely: worktrees, subagents, agent teams, cross-session messaging, and a clean landing routine, step by step. - [Agent-to-agent messaging between coding agents: how it works](https://corkcli.com/blog/agent-to-agent-messaging-coding-agents): How coding agents message each other as of October 2026: Claude Code SendMessage, team mailboxes, cross-session sockets, A2A, and what a useful message holds. - [Messaging between your coding agents with cork message](https://corkcli.com/blog/cork-message-coding-agents): Use cork message to reach every coding agent in a repo, one agent, or whoever is working on a function, across Claude Code, Codex and Cursor. With examples. - [Claude Code agent teams vs separate sessions on one repo](https://corkcli.com/blog/claude-code-agent-teams-vs-separate-sessions): Claude Code agent teams vs separate sessions on one repo: how each works, token cost, same-file conflicts, worktrees, and when to pick which (October 2026). [All articles](https://corkcli.com/blog.md) · [Cork in Markdown](https://corkcli.com/index.md) · [llms.txt](https://corkcli.com/llms.txt) This page in HTML: https://corkcli.com/blog/cork-vs-claude-code-agent-teams