Answers for heavy agent users
Limits, session windows, accounts and keeping your agents running, explained with sources.
-
How to switch Claude accounts without logging out
Three ways to use several Claude accounts in Claude Code without /logout and /login: one config per account, a switcher CLI, or a harness. -
How to install Shuriken on Mac, Windows and Linux
Install Shuriken from shuriken.sh on Mac, Windows or Linux. The Mac app is notarised by Apple; Windows may warn once, as its installer isn't signed yet. -
Can you have multiple Claude accounts?
Yes. Anthropic's terms set no cap on accounts per person, and its phone check allows three per number. Sharing a login and dodging a ban are not allowed. -
Claude's 5-hour limit, explained
Claude's 5-hour window starts with your first message and resets five hours later. An idle account has no window running, so it saves nothing up. -
Claude usage limit reached? What to do next
There's no early reset for Claude's 5-hour or weekly limit. Wait, pay for usage credits, switch to an API key, or carry on with another account you own. -
Claude's weekly limit, explained
Paid Claude plans have a weekly limit on top of the 5-hour window. When it resets, how model limits fit in, and why parallel agents reach it in two days. -
Claude Code 429 error: what it means and how to fix it
A Claude Code 429 rate limit error means you hit an allowance, not an outage: your plan's window, your API tier or a spend cap. Which one, and what to do. -
How to check your Claude Code usage
To check Claude Code usage, type /usage or open Settings › Usage on claude.ai. Both read one pool: your 5-hour session and your weekly limit. -
How to run Claude Code agents in parallel
Run several Claude Code agents on one repo without collisions: a git worktree and branch per agent, merged back one at a time. Commands, limits, pitfalls. -
How to run Claude Code from your phone (2026)
Four ways to drive Claude Code from a phone: Anthropic's Remote Control, the Happy app, SSH with tmux, or a harness. What each needs and what it can't do. -
Claude Pro vs Max for Claude Code
Pro suits a few Claude Code sessions a week; Max suits daily heavy use and parallel agents. Prices, limits, the signs you've outgrown Pro, and upgrading. -
Codex usage limits explained (2026)
How Codex limits work on ChatGPT Plus and Pro: the 5-hour and weekly windows, the 2026 changes, checking yours with /status, and using several accounts. -
Anthropic API rate limit tiers: how to move up
Anthropic API rate limit tiers set a monthly spend cap and per-minute limits. The tiers in October 2026, how to move up, and what Claude Code hits first. -
How to change the model in Claude Code
Change the model in Claude Code with /model, claude --model, an environment variable or settings.json. /model wins; Enter in its picker sets your default. -
Can you use two ChatGPT accounts with Codex? The rules
Multiple ChatGPT accounts for Codex are allowed: OpenAI's help center says you can create more. Sharing a login and circumventing rate limits are not. -
How to choose an app for your coding agents
Picking an app to run Claude Code, Codex and other agents? Eight questions that separate them: where agents run, your plans, worktrees, limits and privacy. -
Claude Code 1M context window: when it helps, what it costs
Opus 4.7+, Sonnet 5+, Haiku 5.5 and Fable run with a 1M context in Claude Code by default. Why a huge session costs more every turn, and when /clear wins. -
Claude Code 529 overloaded: why it happens, what to do
A 529 overloaded error means Anthropic's API is at capacity, not that your quota ran out. Why Claude Code shows it, what helps and what doesn't. -
Claude Code agent teams: what they are and what they cost
Claude Code agent teams run several Claude sessions with a lead, a task list and messages. How to turn them on, what they cost, and when not to. -
Claude Code agent view: run background sessions
Claude Code's agent view (claude agents) lists every background session. They survive a closed terminal and sleep; a reboot stops them until you reply. -
Claude Code on Android with Termux: what works
Claude Code broke on plain Termux in April 2026: it became a glibc binary Android can't run. Ubuntu in proot-distro works; a remote computer works better. -
Claude Code API Error 400 on tool_use: what to do
A Claude Code API Error 400 about tool use means the saved history has a tool call out of step with its result. Update, then rewind past the broken turn. -
Claude Code: API key or subscription?
Pro and Max are flat-priced with 5-hour and weekly limits; an API key bills every token with no plan limits. How to compare with your own numbers. -
Claude Code background tasks: dev servers and Ctrl+B
Ask Claude Code to run a dev server in the background, or press Ctrl+B while it runs. List and stop it with /tasks; exiting Claude Code stops it. -
Claude Code with Amazon Bedrock or Google Vertex AI
Run Claude Code with Amazon Bedrock or Vertex AI: the env vars, the model IDs to pin, IAM, and what works differently from a Claude plan. -
/compact vs /clear in Claude Code: when to use which
Use /clear when you start a new task and /compact when one task has a long history. A decision table, what auto-compact keeps and what a summary loses. -
What Claude Code costs per developer, per month
Claude Code cost per developer per month: Anthropic's average is $150 to $250 at API rates. How it compares with Pro, Max and Team seats, with the maths. -
Claude Code "Credit balance is too low" on a paid plan
On Pro or Max, Claude Code's Credit balance is too low error means it is using an API key with no Console credits, not your plan. Find the key, remove it. -
Claude Code custom slash commands: make your own
Make your own Claude Code slash commands: one Markdown file in .claude/commands or a skill, with arguments, shell output and allowed tools. With examples. -
Claude Code in a dev container: setup, firewall and auth
Claude Code in a devcontainer is one feature line. A named volume and one env variable keep your sign-in through rebuilds; skipping prompts stays unsafe. -
Claude Code down? How to check, and keep working
Claude Code down? Check status.claude.com, /status and /doctor in 30 seconds, tell an outage from your own limit or network, and keep working meanwhile. -
Claude Code thinking and effort levels, explained
Claude Code's effort levels, low to max, set how much Claude thinks before it acts. What each level is for, the defaults, ultrathink, and how to set it. -
Claude Code environment variables: the useful ones
The Claude Code environment variables people actually set, grouped by model, sign-in, network, privacy, updates and behaviour, and where to put them. -
Claude Code for beginners: your first hour
Claude Code for beginners: it needs a paid Claude plan or a Console account, installs with one command, and is safest if you commit to git first. -
Claude Code GitHub Actions: setup with an API key or plan
Claude Code GitHub Actions setup is one command, /install-github-app, and admin access. Runs bill an API key or count against your Claude plan. -
Claude Code in GitHub Codespaces: setup and sign-in
Claude Code runs in GitHub Codespaces through Anthropic's dev container feature, and a personal Codespaces secret keeps each new codespace signed in. -
Claude Code headless mode (-p): scripts, CI and cost caps
Claude Code headless mode is claude -p: one prompt in, one answer out, no terminal UI. The flags that cap spend and tools, output formats, and CI. -
Claude Code chat history deleted? Set cleanupPeriodDays
Claude Code deletes session transcripts after 30 days by default. Raise cleanupPeriodDays in settings.json to keep them, and back up what you can't lose. -
Claude Code hooks examples: 8 to paste into settings.json
Eight Claude Code hooks examples to paste into settings.json: block rm -rf, format on edit, run tests on Stop, desktop alerts, and how exit code 2 works. -
Claude Code hooks not working? Why they don't fire
A Claude Code hook that never fires is either not loaded (missing from /hooks) or loaded with a matcher that matches nothing. The causes and their fixes. -
Claude Code keeps asking you to log in? Fixes
Why Claude Code keeps asking you to log in: expired logins, two programs sharing one login, a stray API key. How to find the cause and fix it for good. -
Claude Code keyboard shortcuts and commands cheat sheet
Every Claude Code keyboard shortcut worth knowing, for Mac and Windows, plus the slash commands that matter. One page, checked against Anthropic's docs. -
Claude Code on a large codebase: what actually works
Claude Code large codebase tips that work: start in the package, a CLAUDE.md per area, block generated code, subagents for reading, tests as the contract. -
Does logging out of Claude Code delete your chats?
No. /logout removes your Claude Code login, not your chats: they stay in ~/.claude/projects. Three other things can hide them or delete them. -
Claude Code MCP server failed to connect: how to fix it
Claude Code says Failed to connect when an MCP server never starts or answers. claude mcp get prints the error; a relative path is a frequent cause. -
Claude Code MCP server timeout: the settings that fix it
Claude Code MCP server timeout errors: the startup timeout for slow servers, a per-server limit for long tool calls, the idle window, and big output. -
Does Claude Code remember between sessions? How memory works
Does Claude Code remember between sessions? Only through CLAUDE.md files and auto memory. A past chat comes back only when you resume it. -
Two Claude Code accounts on Windows (PowerShell, WSL)
Claude Code keeps a Windows sign-in in .credentials.json inside its config folder. One PowerShell function per account gives each its own folder. -
Claude Code across many projects: keep them apart
Claude Code ties sessions, CLAUDE.md, settings and MCP servers to the folder you start it in. Run one session per project; put shared setup in user scope. -
Claude Code with multiple repos: --add-dir and /add-dir
One Claude Code session over multiple repositories: --add-dir gives Claude a second repo's files, not its CLAUDE.md, hooks or MCP servers. The setup. -
Get Claude Code notifications on your phone
Get a phone push when Claude Code finishes or needs approval: Remote Control's built-in push, or a Stop and Notification hook with ntfy. Codex too. -
How to run Claude Code on a Chromebook
Claude Code runs on a Chromebook inside ChromeOS's Linux if the chip is 64-bit, Intel and AMD chips have AVX, and there is 4 GB of RAM. -
Claude Code on a VPS: sizing, setup and staying safe
Claude Code runs on a VPS with 4 GB of RAM, an x64 or ARM64 CPU and Ubuntu 20.04+ or Debian 10+, used from a supported country. How to size and secure it. -
Run Claude Code on an always-on machine
Run Claude Code on a home server, mini PC or VM that never sleeps: install, headless sign-in, tmux, remote access, or add it as a Shuriken server. -
How to use Claude Code on an iPad
Claude Code doesn't run on iPadOS. On an iPad you drive it remotely: Remote Control, cloud sessions at claude.ai/code, or SSH and tmux to your computer. -
Claude Code output token maximum exceeded: the fix
Claude's response exceeded the output token maximum: one reply hit its cap. Raise the max output tokens variable to your model's limit, or split the file. -
Run Claude Code on a remote server over SSH
Run Claude Code over SSH the right way: sign in on a headless box, keep sessions alive with tmux, handle Git keys safely, and see what SSH alone can't do. -
Claude Code permissions: allow, ask and deny rules
Claude Code allowed tools settings explained: the rule syntax, where allow, ask and deny rules live, a safe starter set, and how deny rules protect files. -
Claude Code on a personal account for work: what changes
Claude Code on a personal Pro or Max account at work runs under Consumer Terms: your own training setting and 5-year or 30-day retention apply. -
Claude Code plan mode: when to plan first and when not to
Claude Code plan mode reads your code and writes a plan, but blocks edits until you approve it. Shift+Tab gets you there; skip it for a typo fix. -
Claude Code "Prompt is too long": causes and fixes
Claude Code says Prompt is too long when the request no longer fits the context window. The exact wording of the message points to the cause and the fix. -
Claude Code on a Raspberry Pi: what works
Claude Code runs on a Raspberry Pi 4 or 5 with 4 GB of RAM or more and the 64-bit Raspberry Pi OS. Which boards qualify, the 32-bit trap, and the setup. -
Claude Code Remote Control not working? Fixes
Remote Control fails when an API key or gateway outranks your claude.ai login, an org leaves it off, or the claude process stops. Each message and its fix. -
Claude Code skills vs subagents vs MCP: which one to use
Claude Code skills vs subagents: a skill is know-how in your chat, a subagent is work in its own context, MCP is a live outside system. One table to pick. -
Claude Code status line: show usage, model and cost
The Claude Code status line runs your own script and shows its output at the bottom. It runs locally, uses no tokens, and /statusline writes one for you. -
Claude Code: switch between Team and personal orgs
To switch organization in Claude Code, run /logout, then /login and pick the other org in the browser. /status shows which org and email you're on. -
Claude Code usage analytics for teams: what you can see
The Claude Code analytics dashboard shows a team's users, sessions, lines and PRs. What each plan sees, what OpenTelemetry adds, and what's missing. -
Claude Code on Telegram, Discord or iMessage
Claude Code reaches Telegram, Discord and iMessage through channels, a research preview: an official plugin and claude --channels, on your own computer. -
Claude Code teleport: pull cloud sessions to your terminal
claude --teleport pulls a Claude Code cloud session and its branch into your terminal. It needs a clean tree, the same repo and the same claude.ai account. -
Run multiple Claude Code sessions in tmux
Run several Claude Code sessions side by side in tmux, each in its own git worktree, and keep them running after you close the terminal. Commands included. -
Claude Code in tmux: scrolling, focus and display fixes
Can't scroll Claude Code in tmux? Add set -g mouse on or press PgUp. Copy mode sees history only in the classic renderer. Ctrl+B and colour fixes too. -
Claude Code "Unable to connect to API": proxies and fixes
Claude Code "Unable to connect to API" is usually local: a proxy, a company CA, a VPN or a leftover base URL. One curl in the same shell says which. -
Claude Code using your API key, not your plan? Fix it
Claude Code uses an API key instead of your subscription when one is set, because it outranks your Pro or Max login. Find it with /status and remove it. -
Claude Code in VS Code: switch or split accounts
Switch accounts in the Claude Code VS Code extension with /logout. For an account per project, give a VS Code profile its own Claude config folder. -
Claude Code with VS Code Remote-SSH, WSL and containers
In a VS Code Remote-SSH, WSL or dev container window, the Claude Code extension runs remotely: sign-in, ~/.claude, MCP servers and the CLI live there. -
Claude Code on Windows: native or WSL?
Claude Code runs natively on Windows now. When WSL 2 still wins, the pitfalls that bite on WSL (WSL 1, /mnt/c, sign-in), and how to choose. -
Use Claude Code with GLM or Kimi models: setup and limits
Point Claude Code at GLM or Kimi with the provider's base URL, key and model names. Most of it works; your Claude plan no longer applies. -
How to use Claude Code with Ollama (local models)
Ollama 0.14+ speaks Anthropic's API, so Claude Code runs on a local model via ollama launch claude or two variables. Give the model 64K context first. -
Claude Code --worktree: built-in worktrees, explained
claude -w <name> makes a Claude Code worktree at .claude/worktrees/<name> on branch worktree-<name>. On exit it asks before deleting any work. -
Why Claude Code gets worse in long sessions
Claude Code slips in long sessions because its context window fills, and Anthropic says performance degrades as it does. The signs, and six ways to fix it. -
claude: command not found after installing Claude Code
Claude Code's installer puts claude in ~/.local/bin and leaves your PATH alone. Add that folder to PATH, open a new terminal, and check for a second copy. -
CLAUDE_CONFIG_DIR: two Claude Code profiles on one machine
Point Claude Code at a config folder per profile and keep two accounts signed in at once. One alias each, and what each folder keeps to itself. -
Claude extra usage: how pay-as-you-go overflow works
Claude extra usage, now called usage credits, keeps you working past your plan's limit at API rates. Where to turn it on, the cap, and when it pays. -
Claude Max 5x vs 20x (2026): what you really get
Claude Max 20x gives 20 times Pro's 5-hour allowance and Max 5x five. Anthropic publishes no weekly multiplier; users report 20x at 1.5 to 2 times 5x. -
CLAUDE.md best practices: what to put in, what to leave out
Keep CLAUDE.md short: commands Claude can't guess, rules that differ from defaults, pointers to docs. What to cut, and where must-hold rules go. -
Claude peak hours: do limits really run out faster?
Since March 2026, Claude's 5-hour limits drain faster on weekdays 5–11am PT. Since 6 May, not in Claude Code on Pro or Max. Who's affected, by time zone. -
claude setup-token: the one-year Claude Code token
claude setup-token prints a one-year OAuth token for your Claude plan, for scripts and CI. It runs models only: no Remote Control, no connectors. -
Can you use your Claude plan in OpenCode or other tools?
Not in OpenCode. Anthropic keeps Pro and Max sign-ins for Claude Code and its own apps, and OpenCode dropped its Claude plugins in 1.3.0. Use an API key. -
Claude Team seats for Claude Code, explained
Team Standard seats get 1.25x Pro's usage per session, Premium 6.25x, each with a 5-hour and weekly limit. Prices, Team vs Max, and a personal plan. -
Every Claude usage limit change in 2026, dated
Claude usage limits changed in 2026 more often than the plans did: peak hours, doubled 5-hour limits, a summer weekly boost, and a new weekly level. Dated. -
Codex CLI approval modes and sandbox, explained
Codex CLI approval modes are two settings: the sandbox and the approval policy. Auto pairs workspace-write with on-request; untrusted is retired. -
Codex CLI config.toml: where it lives and what to set
Codex reads ~/.codex/config.toml, a trusted project's .codex/config.toml, a profile and -c flags. Flags win, then the closest project file. -
Codex CLI with local models: --oss, Ollama and LM Studio
codex --oss runs Codex CLI on Ollama or LM Studio, and -m picks the model. A custom provider in config.toml now takes only the Responses wire API. -
How to add an MCP server to Codex CLI
Add an MCP server to Codex CLI with codex mcp add and the server's command, or write its table in config.toml. OAuth servers sign in with codex mcp login. -
Codex CLI on Windows: native PowerShell or WSL?
Codex CLI runs natively in PowerShell inside its own Windows sandbox. In WSL2 it uses the Linux sandbox, needs bubblewrap and no longer runs on WSL1. -
How to switch accounts in Codex CLI without logging out
Switch accounts in Codex CLI without logging out: one Codex home folder per ChatGPT account, with sessions linked in so codex resume finds every chat. -
Codex login not working? Every sign-in error, fixed
Codex login not working? Match your error: port 1455 busy, no browser on a server, refresh token already used, or a config.toml that logs you out. -
Codex MCP server failed to start or timed out: fixes
In Codex, "MCP client failed to start" means the command or handshake broke; "timed out after N seconds" means a slow server: raise its startup timeout. -
Codex remote control: run Codex from your phone
Codex remote control runs through the ChatGPT mobile app and a paired Mac or Windows PC. Linux works only as an SSH host; the CLI command is experimental. -
Codex "stream disconnected before completion": fixes
Codex shows stream disconnected before completion when OpenAI's streamed reply breaks off. It retries 5 times, then stops the turn. How to find the cause. -
Codex worktrees: run several Codex agents in parallel
Codex app: pick Worktree under the composer. Codex CLI: run codex --worktree (0.156.0 or later). Each chat gets its own checkout, so several run at once. -
Multiple accounts in every coding agent CLI: a reference
Multiple accounts in a coding agent CLI: Copilot, Amp and OpenCode 2 switch natively. Most others need a home folder per account, set by one variable. -
Copilot CLI: switch GitHub accounts without logging out
Copilot CLI holds several GitHub accounts: /user list shows them, /user switch changes them. Stuck on the wrong one? A GitHub token in your shell wins. -
How to switch accounts in Cursor, editor and CLI
Cursor keeps one account per install. Sign out in Cursor Settings and sign in again, or run agent logout and agent login. Two at once? A key per account. -
Gemini CLI: change Google account or sign in for work
Change Google account in Gemini CLI with /auth signout, then sign in again. Personal accounts stopped on 18 June 2026; work ones need a Cloud project ID. -
Gemini CLI free tier limits, explained
Gemini CLI's free Google sign-in stopped on 18 June 2026. Code Assist Standard (1,500 requests a day), Enterprise (2,000) and paid API keys still work. -
Gemini CLI resume session: --resume, /resume and history
Resume a Gemini CLI session with gemini --resume, an index or a session ID, or /resume inside it. Where history lives, and why it's gone after 30 days. -
Git worktrees eat disk space: how to keep it under control
Each git worktree uses disk space: a full checkout plus its own dependencies and builds. How to measure it, prune old worktrees and share the heavy parts. -
GitHub Copilot AI credits, explained (2026)
One GitHub Copilot AI credit is $0.01 of model tokens. Pro has 1,500 a month, Pro+ 7,000, Max 20,000. Base credits go first, and all reset on the 1st. -
Install Claude Code on Linux: Ubuntu, Fedora, Alpine
To install Claude Code on Linux, run curl -fsSL https://claude.ai/install.sh | bash, or add Anthropic's signed apt, dnf or apk repository. -
Keep Claude Code running with the MacBook lid closed
caffeinate doesn't survive closing a MacBook's lid. pmset disablesleep does, if you remember to undo it. The free options, the risks, and a one-switch way. -
How to stop Claude Code reading your .env file
Stop Claude Code reading .env with Read(.env) deny rules and the sandbox's denyRead. Anthropic's docs say a .claudeignore file has no effect. -
Keep the same chat when you switch Claude accounts
Claude Code saves each chat in its config folder. How to continue a session on another Claude account by hand, what breaks, and how Shuriken does it. -
How to cap Claude Code spending for a team
Limit Claude Code spending for a team with per-member usage-credit limits, Console workspace limits, a gateway's daily caps or --max-budget-usd per run. -
Set up MCP servers once for every project
Add an MCP server to every Claude Code project with user scope, how .mcp.json and ~/.claude.json differ, and how to keep Codex in step. -
Migrate Claude Code to a new computer, settings and all
To migrate Claude Code to another computer, copy ~/.claude without the login file, keep project paths the same so old chats resume, and sign in again. -
Move Claude Code chat history to a new computer
Copy ~/.claude/projects to the new computer and keep each project at its old path, or rename its folder to match. Then log in, and /resume finds them. -
OpenCode with more than one account per provider
OpenCode multiple accounts: auth.json keeps one login per provider, so a second sign-in replaces the first. A data folder per account keeps both. -
Opus vs Sonnet for coding: which Claude model to use
Opus vs Sonnet for coding: Sonnet 5.5 for well-scoped work, Opus 5.5 for long or open-ended jobs. Opus costs twice as much per token; judge it per task. -
Prompt caching in Claude Code, explained simply
Claude Code caches the start of every request so re-reading your chat is cheap. What is cached, how long it lasts, and what makes it go cold. -
Prompt injection and coding agents: the risks, plainly
Prompt injection turns text a coding agent reads into commands it runs. No model is immune; what limits the damage is what the agent can reach. -
How to write prompts for Claude Code that work first time
How to write prompts for Claude Code: a goal, the files that matter, the limits, and a done-condition it can check. Real before/after prompts. -
Resolving merge conflicts with Claude Code
Parallel Claude Code agents still hit merge conflicts. Merge one branch, rebase the next onto it, give Claude both intents and test before you continue. -
Resume a Claude Code session: --continue, --resume and fixes
Resume a previous Claude Code session with claude --continue in the same folder, or claude --resume <name or ID>. Transcripts sit in ~/.claude/projects. -
How to run Claude Code overnight
Run Claude Code overnight: keep the computer awake, keep the session alive when the terminal closes, and plan for limits and permission prompts. -
"Running scripts is disabled" for codex, claude or gemini
codex.ps1 cannot be loaded because PowerShell blocks the .ps1 shim npm writes. Type codex.cmd, or allow RemoteSigned for your own Windows account. -
Set up Shuriken in your first hour
Set up Shuriken in order: sign in, add a project, add your accounts and keys, choose when chats switch, add MCP servers once, then your phone and servers. -
Is Shuriken worth it with only one Claude account?
With one Claude account, hot swapping has nothing to swap to. What Shuriken still does for you, what it doesn't, and when to stay with the terminal. -
--dangerously-skip-permissions: how to use it safely
Is Claude Code's dangerously-skip-permissions flag safe? Only in a container or VM with no host secrets, scoped keys and a network allow-list. -
Switching from Cursor to Claude Code: what changes
Switch from Cursor to Claude Code: move rules to CLAUDE.md, MCP servers to .mcp.json, and know what you lose and gain. You can keep Cursor open, too. -
Too many MCP servers? How they eat Claude Code's context
Too many MCP servers in Claude Code take little context with tool search on. Tools loaded upfront and big results still fill it, and /context shows which. -
How to undo Claude Code changes with /rewind and git
Undo Claude Code changes with /rewind or Esc twice, and git for the rest. What checkpoints track, the four restore options and the git commands to use. -
How to update Claude Code (and fix a failed update)
Update Claude Code to the latest version: claude update, Homebrew, WinGet, npm or apt. How to tell which install you have, and fixes for failed updates. -
What is an agent harness?
An agent harness is the software that runs a coding agent: sessions, tools, accounts, permissions, devices. How it differs from an orchestrator or IDE. -
Why Claude Code uses so many tokens
Claude Code uses so many tokens because every request resends your whole context: CLAUDE.md, files read, tool output. Caching makes that cheaper, not free. -
Why is Claude Code slow? Causes and fixes
Claude Code is slow for one of three reasons: the model, a huge conversation, or something on your machine. How to tell which, and the fix for each.
Guides are coming soon. Read the docs meanwhile
Try Shuriken on your own computerYou're signed in
Add your Claude, Codex and Gemini accounts once. Shuriken switches between them so your agents keep working.Your accounts, projects and chats are in the app.
Coming soon