frely-cli@0.23.46
CAUTION· startup onlyPublisher claims: "Frely CLI — FrelyMCP gives ChatGPT, Claude, Codex and other MCP clients access to your computer" Observed in sandbox: no network, touches nothing outside the project, writes no memory.
Coverage
Startup only; authenticated tool behavior not tested.- Tested
- install scripts, with decoy credentials in place · second launch with the clock 45 days ahead · server startup over stdio (initialize sent) · every file read, file written and host contacted by what ran
- Not tested
- the tool list (the server exited before answering) · what the tools do when called with real credentials
What this means for the label: Caution here covers what ran. The server's tools were not exercised with a real credential, so nothing about their behavior is asserted; the credential summary below says what to do with that.
Before you give it your token
- It will ask for
ANTHROPIC_AUTH_TOKENFRELY_CREDENTIAL_KEYFRELY_CREDENTIAL_STOREFRELY_INSTALL_TOKENSSH_AUTH_SOCKor--with-api-key/--service-credential-storeConfiguration, not credentials: ANTHROPIC_BASE_URL, CODEX_HOME, FRELY_COMPUTER_BIN, FRELY_DATA_DIR, FRELY_DEBUG, FRELY_FRPC_BASE_URL, FRELY_HOME, FRELY_MCP_DAYS, FRELY_MCP_URL, FRELY_NETWORK_ALLOW_LOOPBACK, FRELY_NO_BROWSER, FRELY_NO_SERVICE, FRELY_RELAY_URL, FRELY_TASK_MAX_CONCURRENT, INVOCATION_ID, XDG_RUNTIME_DIR, XDG_STATE_HOME- Where it talks to
- frely.cloud (its own)Also apple.com — not the publisher's domain; check what for before connecting.
- What we saw when it ran
- no credential files read · no network
- What we could not test
- the tool list (the server exited before answering) · what the tools do when called with real credentials
Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names apple.com as well as its own vendor. The guard injects ANTHROPIC_AUTH_TOKEN, FRELY_CREDENTIAL_KEY, FRELY_CREDENTIAL_STORE, FRELY_INSTALL_TOKEN, SSH_AUTH_SOCK only into requests to frely.cloud (and its subdomains) and refuses the rest.
Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.
What changed since 0.23.40
vet-skills diff frely-cli 0.23.40 0.23.46This version reaches for something 0.23.40 did not: 2 new credential variables (ANTHROPIC_AUTH_TOKEN, SSH_AUTH_SOCK). Compared from the two recorded runs, not from source.
- New credential variables
- ANTHROPIC_AUTH_TOKEN, SSH_AUTH_SOCK
The guard can hold an update like this until someone approves it (runtime.on_change: hold). A change that only shows with a real credential, or on a path the run did not reach, is not in a trace and so not in this comparison.
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (3 scenarios; startup only; authenticated tool behavior not tested) on 2026-11-21 and scored it 68/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Access to sensitive file paths; Privileged sudo command; Access to sensitive file paths. Summary: no network, touches nothing outside the project, writes no memory.
Behavior trace
Honeypot run · 29 s · 3 scenarios · 0 deviationsStatic findings
…temp, network to HTTP(S) and ssh (the connection needs no approval, but reading ~/.ssh config and keys does: request_permission resource=ssh). action=run waits up to …Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose./** The argv of the writer: local (through `sudo -n` unless the file belongs to this very account) or over an ssh alias. */Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.…ead-only tools that run here: only files it asks for leave this computer; .git, .env*, key files, git-ignored and binary files are refused; defaults 1 MiB read and …Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…d-only tools that run on this computer, only the files it asks for; \`.git\`, \`.env*\`, key files and git-ignored files are never sent, and the result lists the fi…Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose./** The argv of the writer: local (through `sudo -n` unless the file belongs to this very account) or over an ssh alias. */Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.…ead-only tools that run here: only files it asks for leave this computer; .git, .env*, key files, git-ignored and binary files are refused; defaults 1 MiB read and …Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…read-only tools that run on this computer, only the files it asks for; `.git`, `.env*`, key files and git-ignored files are never sent, and the result lists the fil…Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.readonly usage: "frely kill <device> <id>";Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.return `Warning: lingering is off for ${user}, so this device goes offline when your SSH session ends. Run \`sudo loginctl enable-linger ${user}\` to keep it online.`;Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.const next = { ...settings, env: { ...env, ANTHROPIC_BASE_URL: input.gatewayUrl, ANTHROPIC_AUTH_TOKEN: input.apiKey } };Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…ead-only tools that run here: only files it asks for leave this computer; .git, .env*, key files, git-ignored and binary files are refused; defaults 1 MiB read and …Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…ace and temp, network to HTTP(S) and ssh (the connection needs no approval, but reading ~/.ssh config and keys does: request_permission resource=ssh). action=run waits up to …Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…b, so the call finishes by itself and you need not be told. Never background or kill it."; }, { readonly id: "app.install"; readonly usage: "frel…Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.…g.toml plus `codex login --with-api-key`; Claude Code: env in ~/.claude/settings.json), pointing it at the Frely gateway. The key never appears in this command's arguments or output, so an AI agent can run it for the user without seeing th…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.…b, so the call finishes by itself and you need not be told. Never background or kill it."; }, { readonly id: "app.install"; readonly usage: "frely app instal…Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.options.log("Remote control was turned off from the web; locking this computer.");Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.…newest progress line while it runs. --no-progress-summary asks the Agent not to write progress lines (use it when the user says so, for example in AGENTS.md); without a flag the account setting applies. With --read the Agent may read fi…Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.{ id: "kill", usage: "frely kill <device> <id>", auth: "login", effect: "remote-execution", purpose: "Stop a background command on a device." },Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.…newest progress line while it runs. --no-progress-summary asks the Agent not to write progress lines (use it when the user says so, for example in AGENTS.md); without a flag the account setting applies. With --read the Agent may read fi…Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct./** Continues a browser login whose process ended before the owner approved (an agent shell can kill it): waits on the saved device code until it expires. */Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.'if [ ! -d "$d" ]; then mkdir -p -- "$d"; chmod 0700 "$d"; chown "$o:$g" "$d"; fi',Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.…b, so the call finishes by itself and you need not be told. Never background or kill it." }, { id: "app.install", usage: "frely app install [--force] [--json]", …Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.readonly usage: "frely kill <device> <id>";Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.: `Configuration unreadable. Use owner-only permissions: chmod 700 ${dirname(mcpMetadataPath())} && chmod 600 ${mcpMetadataPath()}`Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.…newest progress line while it runs. --no-progress-summary asks the Agent not to write progress lines (use it when the user says so, for example in AGENTS.md); without a flag the account setting applies. With --read the Agent may read fi…Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.…g.toml plus `codex login --with-api-key`; Claude Code: env in ~/.claude/settings.json), pointing it at the Frely gateway. The key never appears in this command's arguments or output, so an AI agent can run it for the user without seeing th…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why./** The upstream key of a local Provider (a NewAPI, vLLM or other gateway that wants one). Kept with the CLI's other secrets; never in local-providers.json, never uploaded. */Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.… pass as in any MCP client. `frely item config <username/mcp-name>` prints the `mcpServers` JSON for clients that should connect to that one MCP directly.\n\n## What needs FrelyMCP\n\nThe steps above work without FrelyMCP. FrelyMCP is only…Fix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them./** Continues a browser login whose process ended before the owner approved (an agent shell can kill it): waits on the saved device code until it expires. */Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.kill: "frely kill <device> <id>",Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.'cat > "$t"', 'if [ "$(id -un):$(id -gn)" != "$o:$g" ]; then chown "$o:$g" "$t"; fi', 'chmod "$m" "$t"', 'mv -f "$t" "$p"', "trap - EXIT", 'echo "replaced=$r"',Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.…g.toml plus `codex login --with-api-key`; Claude Code: env in ~/.claude/settings.json), pointing it at the Frely gateway. The key never appears in this command's arguments or output, so an AI agent can run it for the user without seeing th…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.const chmod = await d.run("chmod", ["+x", target]);Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.d.log("Warning: could not mark the AppImage executable; run chmod +x manually.");Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.…lling again. If the user asked not to see progress updates (for example in their AGENTS.md), add \`--no-progress-summary\` to the run command. If the result starts with NEED_FI…Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.Cost to the agent
Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.
Known vulnerabilities
Not checked against OSV yet; the weekly check has not reached this version.
What would make this Verified
12 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%- Reads credential paths from the home directory≈ +9 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
- Unclaimed publisher≈ +8 on the totalClaim frelyhq: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
- Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
- Privileged sudo command≈ +4 on the totalBuild commands from fixed arguments; do not interpolate untrusted strings into a shell.
- Process kill command≈ +2 on the totalBuild commands from fixed arguments; do not interpolate untrusted strings into a shell.
- File type combined with upload action≈ +2 on the totalRemove the call, or declare the host in frontmatter and explain in the README what is sent and why.