# AI use cases for Agentreach (bojieli/agentreach) Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Use cases: 5 | Page: https://dedupedai.com/use-cases/bojieli-agentreach/ AgentReach (reach) lets a coding agent running on your machine execute its shell commands and file operations on a remote box you can reach over SSH, without installing the agent or your API key on that box. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run Claude Code against a remote box without installing it there: https://dedupedai.com/use-cases/bojieli-agentreach/run-claude-code-on-remote-box/ - Let Codex work on a client's server with no key left behind: https://dedupedai.com/use-cases/bojieli-agentreach/drive-codex-on-client-server/ - Run remote tests from a throwaway container or CI job: https://dedupedai.com/use-cases/bojieli-agentreach/run-tests-from-disposable-container/ - Audit what an agent ran or changed on a server: https://dedupedai.com/use-cases/bojieli-agentreach/audit-agent-changes-on-server/ - Keep several remote machines open in parallel with one session each: https://dedupedai.com/use-cases/bojieli-agentreach/manage-several-remote-boxes-at-once/ ## How to run Claude Code against a remote box without installing it there with Agentreach Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Point Claude Code at an SSH host so its shell and file work run on that host. **Who it is for:** A developer who needs to diagnose or fix something on a build box or VM and wants an agent to help, but does not want a 300 MB agent install or an API key on a machine other people can access. **The problem:** Installing a coding agent on a remote machine puts its binary and API key on a box you do not own or will not keep, and the key stays behind after you leave. **The outcome:** The agent runs locally, its commands and file reads or writes happen on the target, and nothing is left on the target after the session closes. ### Steps 1. Install reach with `go install github.com/bojieli/agentreach/cmd/reach@latest` 2. Start the session with `reach build-box claude`, using your ssh_config alias or a host:path target 3. Ask the agent a question about the box, for example what is using disk space, and let it run commands 4. Review what it did with `reach log` 5. Close the session with `reach down build-box-app` or the session name shown by `reach status` ### Prompt to paste into your agent ```text Use the repo bojieli/agentreach to run Claude Code against , for example build-box:/srv/app. Ask it what is using disk space on the target and confirm the commands ran there by checking `reach log`. Expected output: a disk usage summary from the target, and log entries showing exec commands with status ok. ``` ### Caveat Claude Code is the harness the README names for mirror mode, and reach relies on the ssh binary you already have, so ProxyJump, IdentityFile and 2FA behave as they do in your ssh config. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/bojieli-agentreach/run-claude-code-on-remote-box/ ## How to let Codex work on a client's server with no key left behind with Agentreach Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Run Codex against a remote project directory so its tools act on that server. **Who it is for:** A contractor or consultant who must change code or config on a client's server and does not want an agent install or API key left on a machine they do not control. **The problem:** An agent installed on a client server keeps the operator's API key in env files and shell history, and that key is exposed if the server is compromised. **The outcome:** Codex edits and runs commands on the client's project directory while the key stays on your own machine. ### Steps 1. Install reach with `go install github.com/bojieli/agentreach/cmd/reach@latest` 2. Start the session with `reach client-box:/srv/app codex` 3. Ask Codex to make the change and let it run its commands on the server 4. Check the record of commands and writes with `reach log` 5. Close the session with `reach down` using the session name from `reach status` ### Prompt to paste into your agent ```text Use the repo bojieli/agentreach to run Codex on , for example client-box:/srv/app. Ask it to list the project files and report the current config value you name, then confirm the work ran remotely with `reach log`. Expected output: the file listing and config value from the server, plus log entries for the exec and write actions. ``` ### Caveat Codex is described as the clean fit, but the README says it has been verified end-to-end only at version 0.148.0, so other versions may behave differently. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/bojieli-agentreach/drive-codex-on-client-server/ ## How to run remote tests from a throwaway container or CI job with Agentreach Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Use a disposable container with reach to run commands on a target box. **Who it is for:** A developer or CI maintainer whose agent or test job runs in a short-lived container and must run commands on a build box. **The problem:** Baking an agent into every dev container bloats the image, and mounting one from the host breaks easily. **The outcome:** The disposable container runs reach against a target over SSH and returns command output and exit codes, with no agent or extra credentials baked into the image. ### Steps 1. Run `docker run --rm -v "$HOME/.ssh:/root/.ssh:ro" ghcr.io/bojieli/agentreach version` to confirm the image works 2. Mount your keys and known_hosts into the container as shown in the command above 3. Run a command on the target with `reach build-box exec -- go test ./...` 4. Check the target with `reach build-box doctor` 5. Remove the container when the job finishes, since it leaves no state behind ### Prompt to paste into your agent ```text Use the repo bojieli/agentreach from its container image to run on through `reach exec -- `. Expected output: the command's stdout and an exit code. Verify that `reach doctor` reports the target platform before the run. ``` ### Caveat The README says the image carries helper binaries for every target platform but that reach will not invent credentials, so the mounted keys must already work for the target. Risk, credentials: Your SSH keys and known_hosts must be mounted into the container, so the container can use them while it runs. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/bojieli-agentreach/run-tests-from-disposable-container/ ## How to audit what an agent ran or changed on a server with Agentreach Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README Use the reach log to review every command and write an agent made on a target. **Who it is for:** An ops engineer who let an agent work on a server and needs to confirm exactly which commands ran and which files changed. **The problem:** After an agent session on a remote machine, it is hard to tell which commands ran and which files were rewritten. **The outcome:** A timestamped list of commands, exit codes and file writes that reach made on the target. ### Steps 1. Start or continue a session with `reach build-box claude` 2. Let the agent work on the target 3. Review the record with `reach log` 4. Compare any listed write against the file on the target 5. Close the session with `reach down` when finished ### Prompt to paste into your agent ```text Use the repo bojieli/agentreach to review the actions taken on during an agent session. Run `reach log` and report each exec and write entry with its status. Expected output: a table with WHEN, ACTION, STATUS and DETAIL columns that matches what the agent did. ``` ### Caveat The README describes reach log as a record of what reach ran and changed, so commands run outside reach are not part of this audit. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/bojieli-agentreach/audit-agent-changes-on-server/ ## How to keep several remote machines open in parallel with one session each with Agentreach Repo: [bojieli/agentreach](https://github.com/bojieli/agentreach) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Open one reach session per box and track them with reach status. **Who it is for:** An engineer juggling several build boxes, staging VMs or client servers who wants a separate agent session for each one. **The problem:** Keeping track of which agent session belongs to which machine gets messy as the number of boxes grows. **The outcome:** Each target has its own named session, and reach status lists them with their mode, file ops and working directory. ### Steps 1. Open a session for each target with `reach build-box:/srv/app claude` (inferred, check the README) 2. Open another session with `reach client-box:/srv/app codex` 3. List the active sessions with `reach status` 4. Reuse a session by running another command against the same target 5. Close each one with `reach down build-box-app` using the name from the status list ### Prompt to paste into your agent ```text Use the repo bojieli/agentreach to list the active sessions on this machine with `reach status`, using as a reference. Expected output: a table with NAME, TARGET, MODE, FILEOPS and CWD for each open session. Verify that a second command against the same target reuses its session. ``` ### Caveat Sessions are named after their target by default, so two targets with the same host name would need a chosen name from reach up --name. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/bojieli-agentreach/manage-several-remote-boxes-at-once/