DEV Community: AgentVera Team

MCP server managers for Claude Code, Codex and Gemini CLI (2026)

AgentVera Team — Tue, 03 Nov 2026 03:08:54 +0000

Short answer: If you run Claude Code, Codex and Gemini CLI in your own terminals and want one MCP server list for all of them, use a desktop manager that writes each client’s config: Toolport (MIT, with keychain secrets and paid sync across machines) or MCP Manager by xjeway (MIT, with a registry browser). Codeg, an open-source multi-agent workspace, does the same for the agents it runs and installs servers from the official MCP Registry and Smithery. If you run your agents inside one app, AgentVera’s MCP manager passes the same servers to Claude Code, Codex, Gemini CLI and 14 more CLI agents when they start, with one sign-in for remote servers.

We build AgentVera, one of the tools below. We have described every tool, ours included, from its own site, docs or repository as of 3 November 2026. If you only use Claude Code, our Claude Code MCP setup guide covers the built-in commands, and the best MCP servers for coding covers what to install.

Three ways to keep one MCP list

Write every client’s config. The manager keeps the master list and writes it into ~/.claude.json, Codex’s config.toml, ~/.gemini/settings.json, ~/.cursor/mcp.json and so on. Codeg and xjeway’s MCP Manager work this way.
Point every client at one hub. Each client gets a single entry that leads to a local gateway, and the gateway runs the real servers. Toolport and vietqtran’s MCP Manager work this way.
Hand servers over at launch. The app that starts the agent passes the servers on the command line, so the client’s own file stays as it was. AgentVera works this way for the agents it runs.

Built-in settings in Claude Desktop, Cursor and VS Code are a fourth option, but each one manages only its own client.

MCP server managers compared

Checked on each tool’s own site, docs or repository on 3 November 2026.

Tool	Clients supported	Writes each client’s config	Platforms	License	Price	Extra
Toolport (toolport.app)	Claude Code, Claude Desktop, Codex, Cursor, VS Code, Gemini CLI, Zed, Cline, OpenCode and more, 45 in all	Yes: replaces each agent’s direct entries with one Toolport gateway entry, with a backup before every edit	Windows, macOS, Linux	MIT	Free on one machine; Pro $5/month or $48/year; Team $19/month for up to 10 people; self-host free for up to 10	OS keychain for keys; Pro adds 1Password, Doppler, Infisical, Vault and sync across machines; Team adds roles, audit log, change review; no server catalog described
Codeg (codeg.app)	18 of its 20 built-in agents, incl. Claude Code, Codex, Gemini, OpenCode, Cline, Cursor, Kiro CLI	Yes, into each agent’s native config; removes it when you untick	Desktop on macOS, Windows, Linux; headless server and Docker; iOS and Android clients	Apache-2.0	No price stated	Market tab with the official MCP Registry and Smithery; scans existing configs; keys stored in agent configs, not encrypted at rest
MCP Manager (github.com/xjeway/mcp-manager)	24, incl. Claude Code, Claude Desktop, Codex, Cursor, VS Code, Gemini CLI, Copilot, Zed, Goose, Kiro	Yes: imports from each client and applies the list back, with review of risky writes, backups and rollback	macOS, Windows, Linux; plus a preview CLI	MIT	No price stated	Marketplace over the GitHub MCP Registry and the official MCP Registry; secrets and sync not described
MCP Manager (github.com/vietqtran/mcp-manager)	Claude Code, Codex, Cursor, VS Code, Gemini CLI, Claude Desktop (via a bridge)	One hub entry per client, written in its own format, with a backup	macOS 14+ (build from source)	MIT	No price stated	One background engine on one HTTP endpoint; per-tool hiding; auto-restart; secrets in a user-only file; 36 presets
MCPM (mcpm.sh)	Claude Desktop, Cursor, Windsurf, VS Code, Cline, Continue, Goose, Roo Code, OpenCode and others; Claude Code, Codex and Gemini CLI are not in its client list	Enables or disables servers per client from the CLI	Command line, installed with Homebrew, pipx, uv or pip	MIT	Free	Profiles; its own registry at mcpm.sh/registry; sharing over a tunnel
Claude Desktop settings	Claude Desktop only	Its own claude_desktop_config.json (Settings, Developer, Edit Config)	macOS, Windows, Linux	Proprietary	Part of the Claude app	Connectors menu to see each server’s tools
Cursor MCP settings	Cursor only	Its own .cursor/mcp.json and ~/.cursor/mcp.json	macOS, Windows, Linux	Proprietary	Part of Cursor	One-click installs from the Cursor Marketplace and cursor.directory; OAuth; ${env:NAME} for secrets; team admins can share servers
VS Code MCP settings	VS Code and Copilot	Its own mcp.json files and Copilot’s mcp-config.json	macOS, Windows, Linux	Proprietary build of MIT-licensed source	Free	MCP server gallery in the Extensions view; organisations can restrict servers
AgentVera MCP manager	Claude Code, Codex, Gemini CLI and 14 more CLI agents (all but Aider)	Passes servers at launch and leaves your files alone; can import from or write to Claude Code’s and Codex’s settings	macOS, Windows, Linux	Proprietary	Free plan	23-server catalog; OS keychain for secrets; one OAuth sign-in for remote servers through a gateway; connection test with token cost per tool; per-project, group and role servers

Notes: MCP Router, another desktop manager, ended development on 18 September 2026 and its repository is archived, so it is not in the table. vietqtran’s MCP Manager has no published release yet; its README describes building from source. This space changes fast, so check each project before you decide.

Which one fits?

You run agents in your own terminals

Pick a manager that writes the configs, so the servers are there whichever terminal you open. Toolport suits you if you want a polished app with keys in the keychain and, later, the same setup on several machines or for a team. xjeway’s MCP Manager suits you if you want a free, MIT app that browses MCP registries and applies one list to 24 clients with a review step. vietqtran’s MCP Manager suits a Mac user who prefers one shared engine and one entry per client over copies in every file.

You want one workspace for the agents too

Codeg and AgentVera both run several coding agents side by side, so MCP management is part of a larger app. Codeg writes servers into each agent’s native config and installs from two public registries; it is open source and also runs as a server with mobile clients. AgentVera hands servers to each agent at launch and adds remote-server sign-in, token cost per tool and servers per agent group.

You use one client

Use its built-in settings. Claude Desktop, Cursor and VS Code all have a server list, and Cursor and VS Code have one-click installs. For Claude Code on its own, claude mcp add is enough; see our setup guide.

Where secrets live

If API keys in plain JSON or TOML files worry you, look for keychain storage (Toolport, AgentVera), a secrets manager (Toolport Pro) or environment variable references (Cursor). Codeg’s docs say plainly that keys written into agent configs are not encrypted at rest.

How AgentVera’s MCP manager works

AgentVera is a desktop app for macOS, Windows and Linux that runs coding agents side by side in real terminals. Its MCP manager keeps one server list for all of them:

Add a server once. On the MCP page (⌘⌥M), “+” lists 23 ready servers by category, such as Playwright, Chrome DevTools, Context7, GitHub, Sentry, Supabase, Linear, Notion, Stripe and Figma. You can also add your own stdio, HTTP or SSE server.
Every CLI agent gets it. Registered servers connect when an agent starts: Claude Code, Codex, Gemini CLI, opencode, Copilot, Qwen Code, Goose, Amp, Cursor, Antigravity, Droid, Kimi, Vibe, Crush, Kiro, Cline and Hermes. Aider takes no MCP. You choose all agents, some agents, a group or a team role.
One sign-in for remote servers. Remote servers go through AgentVera’s gateway. You sign in to an OAuth server once in the app and every agent uses that sign-in; the agent’s CLI holds no header or token.
Test it, see the cost. A connection test lists the server’s tools, prompts and resources and an approximate token cost per tool, so you know how much context a server adds. Secret environment variables and headers are kept in the operating system’s keychain, and servers can be limited to certain projects.
In step with your CLIs. Servers from Claude Code’s and Codex’s own settings are listed with their health; import them or remove them from the CLI. AgentVera’s servers can be written to the CLI settings too.
Agents ask for servers. An agent can attach a registered server to itself or to an agent of its group. A new server is only added when you approve it in the app.

What to weigh: AgentVera is not open source. The MCP manager hands servers to agents that AgentVera starts; a claude you open in another terminal sees only what is in its own settings, unless you write AgentVera’s servers there. The MCP manager is on the free plan; see pricing.

Originally published on the AgentVera blog.

Claude Code account switchers compared: auto-switch at the limit

AgentVera Team — Tue, 03 Nov 2026 01:22:27 +0000

Short answer: yes. If Claude Code keeps hitting the usage limit and you have more than one account you are entitled to use, these tools move you to the next account automatically: ccswap and claude-swap in the terminal, AgentPool inside VS Code, ccam (opt-in, experimental) and AgentVera in a desktop app. ccswap, AgentPool and AgentVera also switch Codex accounts. Most of them can keep the same conversation going on the new account.

We build AgentVera, one of the tools below. We describe every tool, ours included, by what its own site, repository or marketplace page said on 3 November 2026.

If you only want to run two accounts side by side by hand, you don’t need any of these. One config folder per account does it; see how to switch Claude Code accounts without logging out. This post is about the next step: switching automatically when a limit is near.

Automatic account switchers compared

Tool	Agents	Platforms	Auto-switch at the limit	Same conversation after a switch	License	Price
ccam	Claude Code	Linux, macOS, Windows	Opt-in and experimental (`auto_switch`, Linux and macOS); otherwise you run `ccam next --auto`	Yes, with `ccam next --continue`	MIT	Free
AgentPool (VS Code extension)	Claude Code, Codex	Wherever VS Code runs: macOS, Windows, Linux	Yes, before the limit (default: 95% of the 5-hour window used), and back when a higher-priority account has room	Yes; before a switch to another organisation it asks you to run `/compact`	MIT	Free
ccswap	Claude Code, Codex	macOS, Linux, Windows	Yes, `ccswap auto` checks every 60 s and switches at 90% used by default	Claude Code usually without a restart; Codex needs a restart	MIT	Free
claude-swap (`cswap`)	Claude Code (CLI and VS Code extension)	macOS, Linux, Windows	Yes, `cswap auto` checks every 60 s, 90% used by default	Yes, the same session keeps going	MIT	Free
codex-swap	Codex	macOS only	Picks the least-used account each time you launch Codex; no mid-session switch documented	Not documented	MIT	Free
AgentVera	Claude Code, Codex (chains); per-agent accounts for other CLIs	macOS, Windows, Linux	Yes, account chains (default: 3% left)	Yes, when “Move the conversation to the new account” is on	Closed source	Free app; chains on Pro and Team

Checked on 3 November 2026 against each tool’s own page: ccam.dev and its repository, the AgentPool listing on the VS Code Marketplace, ccswap, claude-swap and codex-swap on GitHub. These tools change every few weeks, so check before you install.

Two notes on names. ccswap began as a fork of claude-swap and added Codex; the two are now separate projects. And the AgentPool in this table is the VS Code extension; a different project, agentpool-cli on PyPI, reads limits across several agents but leaves the choice of provider to you.

All of these tools read quota from usage endpoints that Anthropic and OpenAI don’t publish as a stable API (ccam’s README says so plainly). Treat the percentages as a guide, not a promise.

Which one fits you?

ccam if you use Claude Code in a terminal and want something small

ccam is a single Go binary. Each account gets its own config folder, and ccam <name> or a shortcut like ccwork opens Claude Code as that account. ccam next --auto --continue moves to the account with the most room and keeps the conversation. Fully automatic switching (auto_switch = true) is opt-in and marked experimental, on Linux and macOS.

AgentPool if you work in VS Code

AgentPool lives in the VS Code status bar. Hovering shows every Claude Code and Codex account’s 5-hour and weekly quota, you drag accounts into a priority order, and auto switch walks that order before the account in use runs out. Account groups keep company projects on company accounts and personal projects on personal ones. It changes the login of Claude Code inside VS Code; the claude in your own terminal keeps its account unless you turn on a setting.

ccswap if you use both Claude Code and Codex from the terminal

ccswap keeps several Claude Code and Codex logins, shows their usage in a dashboard and switches with ccswap auto before a limit hits. Claude Code usually continues without a restart. Codex reads its login when it starts, so a running Codex session has to be restarted.

claude-swap if you only use Claude Code

claude-swap (cswap) is the project ccswap was forked from. It covers Claude Code only, works with the CLI and the VS Code extension, and has strategies such as “most quota left” or “use up the account that resets first”, with a cooldown so it doesn’t switch back and forth.

codex-swap if you use Codex on a Mac

codex-swap wraps codex with a cx command that launches Codex on the ChatGPT account with the lowest usage. It decides at launch, not mid-session, and runs on macOS only.

AgentVera if you run several agents at once

The tools above switch one login: the one your terminal or editor is using. AgentVera is a desktop app that runs many Claude Code, Codex and other CLI agents side by side, and each agent has its own account. An account chain moves only the agents whose account is running out, one by one, and can take each agent’s conversation with it. That matters when one agent is on a work account, another on a personal one, and only one of them is near its limit.

What to weigh: AgentVera is closed source, and automatic switching is on the Pro and Team plans. If you use one Claude Code session in one terminal, a free terminal tool may be all you need.

How account chains work in AgentVera

This is what the desktop app does, read from its source. Setting a chain up step by step is in the account switching guide; here is what happens once it runs.

Every agent has its own account

Each extra Claude Code or Codex account gets its own config folder, so several accounts stay signed in at once, and you pick an account for each agent. The account list shows who each one is signed in as: email, organisation and plan, since the same email can be on a personal plan and on a team plan. Other CLIs such as Cursor, Gemini CLI, Copilot or OpenCode can have extra accounts too, with their own sign-in folder or with API keys, chosen per agent. Automatic chains cover Claude Code and Codex.

The check every 90 seconds

While a chain is on, AgentVera reads the 5-hour and weekly usage of every account in it every 90 seconds. An account counts as out of quota when either window has the threshold or less left; the default is 3%, adjustable up to 30%. If the CLI itself prints a limit message, AgentVera reacts right away instead of waiting for the next check.

A running turn is allowed to finish

When an account is only near its threshold and the agent is in the middle of a turn, the switch waits for the turn to end, so the agent isn’t cut off mid-edit. It waits at most 20 minutes; a turn that runs longer is switched anyway. If the CLI has already stopped on the limit, there is nothing to wait for and the switch happens at once.

A 5-minute cooldown

An agent that was just switched isn’t switched again for 5 minutes, so a flickering usage reading can’t bounce it between accounts. The exception is when the account it just moved to turns out to be out of quota too; then it goes on along the chain without waiting.

Picking the next account

AgentVera takes the next account after the current one in the chain and wraps around at the end. It skips accounts that are out of quota, signed out, or whose stored login the provider rejects. An account that hit its limit stays skipped until its window resets (an hour, if the reset time isn’t known). If no account is left, AgentVera warns you, and when the limit resets the agent is told to continue on its own account.

The conversation comes along

With “Move the conversation to the new account” on, the conversation is copied to the new account’s folder and the agent restarts on it. If the agent had unfinished work, AgentVera types a short note into it: the previous account hit its usage limit, it is on the next account with the same conversation, and it should continue where it left off. Agents running on SSH servers use the server’s own login and aren’t switched.

A Pro feature

Account chains are a Pro feature (also in Team), available in the released desktop app. On the free plan you can still add accounts and choose one per agent; when an agent hits its limit, AgentVera tells you instead of switching. Plans are on the pricing page.

Use only accounts you are entitled to

An account switcher is a convenience for accounts you already have a right to use. It is not a way around the limits you agreed to.

Good reasons to have several accounts: a personal plan and a work account, a team seat from your employer next to your own subscription, or separate accounts that clients pay for.
Read the terms: check Anthropic’s terms and OpenAI’s terms for the plans you use, and your employer’s policy for work accounts. Whether a setup is allowed depends on them, not on any tool here.
Don’t share or borrow logins, and don’t create accounts only to stack quota. ccam’s own site makes the same point: using more than one account depends on the terms of each account.

If you mostly want the quota you have to last longer, start with Claude Code usage limits: smaller context, the right model for the job and fewer wasted turns go a long way before a second account is needed.

Originally published on the AgentVera blog.

Vibe Kanban alternatives after the Bloop shutdown (2026)

AgentVera Team — Mon, 02 Nov 2026 23:36:58 +0000

Short answer: Vibe Kanban still works as a local, open-source tool, but its company (Bloop) shut down on 10 April 2026 and the cloud boards are gone. If you only ran agents on your own machine, you can keep it or move to Claude Squad or Nimbalyst. If you need a shared board that hands tasks to agents, look at AgentVera. If you are on a Mac and want a polished app, look at Conductor.

We build AgentVera, one of the options below. We have tried to describe every tool, ours included, by what its own site and repository say as of November 2026.

What happened to Vibe Kanban?

Vibe Kanban was built by Bloop: a kanban board where each card runs a coding agent (Claude Code, Codex, Gemini CLI, Amp and others) in its own git worktree, with a web UI for reviewing the diff.

On 10 April 2026 Bloop announced it was shutting down. The repository now opens with the line “Vibe Kanban is sunsetting”. In practice:

The code stays open source (Apache-2.0, about 28k GitHub stars) and community maintainers still commit to it.
Remote services ran for 30 more days; after that the app became fully local.
Cloud features were removed: shared kanban issues, comments, projects and organisations. Bloop offered a data export.
Releases slowed down: the last tagged release as of November 2026 is v0.1.44 (24 April 2026).

So the question depends on how you used it. Solo and local: little has changed. Team boards: you need something else.

Vibe Kanban alternatives compared

Tool	Runs on	Agents	Interface	License	Price
Vibe Kanban (community)	Local CLI + browser	~10, incl. Claude Code, Codex, Gemini CLI, Amp	Kanban board	Apache-2.0	Free
AgentVera	macOS, Windows, Linux	18 CLI agents	Desktop app: terminal grid + team kanban	Closed source	Free; kanban on Pro/Team
Nimbalyst (formerly Crystal)	macOS, Windows, Linux	Claude Code, Codex, OpenCode	Desktop app, mobile companion	MIT	Free for individuals
Paradev (formerly Superset)	macOS (Linux experimental)	Several CLIs	Desktop app	Elastic License 2.0	See its site
Conductor	macOS only	Claude Code, Codex, Cursor	Desktop app	Closed source	See its site
Claude Squad	macOS, Linux (Windows via WSL)	Claude Code, Codex, Gemini, Aider	Terminal (tmux)	AGPL-3.0	Free

Facts are from each tool’s own site and repository as of November 2026; this space changes quickly, so check before you decide.

Which one fits you?

Keep Vibe Kanban if you worked alone and locally

The board, the worktree per task and the agent support you used are still there. Pin the version you use, keep an eye on the repository, and you lose little. The risk is slower fixes as agent CLIs change.

Claude Squad if you live in the terminal

Claude Squad runs each agent in its own tmux session and git worktree, in a terminal UI. It is AGPL-3.0, light and scriptable. There is no board; you switch between sessions from a list. On Windows it needs WSL because it relies on tmux.

Nimbalyst if you want an open-source desktop app

Nimbalyst is the successor of Crystal (deprecated in February 2026). It runs Claude Code, Codex and OpenCode sessions in worktrees, works on all three desktop systems, has a mobile companion and is free for individuals under the MIT license.

Conductor if you are on a Mac and use Claude Code or Codex

Conductor is a polished macOS app that gives each agent its own workspace and turns finished work into a pull request. It does not run on Windows or Linux; we covered that in Conductor on Windows.

Paradev if source-available is enough

Paradev (formerly Superset) is a desktop app for running several CLI agents in parallel. Its code is public under the Elastic License 2.0, which is source-available rather than open source.

AgentVera if you need a team board that hands cards to agents

This is the part of Vibe Kanban the shutdown took away, and it is what AgentVera’s kanban for AI agents does:

A board shared with your organisation. You send a card to Claude, Codex or another CLI agent on your own computer; the card moves to In progress and, when the agent’s turn ends, its last reply is posted as a comment.
Only cards, comments and assignments go to the cloud. Your project folder stays on your machine.
Agents run in real terminals, side by side, each with the CLI login you already have (Claude Pro/Max, a ChatGPT plan and so on). No API key is required.
You can create a git worktree per task from the Git panel, like a Vibe Kanban card.

What to weigh: AgentVera is not open source, and the kanban board is on the Pro plan (10 boards) and the Team plan (unlimited). The grid, worktrees and all 18 agents are on the free plan. A side-by-side comparison is on the Vibe Kanban vs AgentVera page.

Moving off Vibe Kanban

Export what you need from the old cloud boards if you still have the export; Vibe Kanban no longer syncs them.
Finish or merge open worktrees. Vibe Kanban’s task worktrees are ordinary git worktrees: git worktree list shows them, and any tool above can open them.
Keep your agent logins. Every tool here runs the same CLIs you already installed, so there is nothing to re-buy.
Recreate only the open cards. Done cards are history; start the new board with what is still in flight.

For a broader look at tools that run agents in parallel, see the best tools for parallel AI coding agents.

Originally published on the AgentVera blog.