Root cause: the launch command did `source ~/.bashrc; source ~/.zshrc; claude`.
The .zshrc contains `eval "$(fnm env)"` which outputs PATH with literal
"$PATH" in quotes instead of expanding it, destroying the entire PATH.
Confirmed via debugging:
- `ssh -t ... 'export PATH=...; which claude'` → works (/root/.bun/bin/claude)
- `ssh -t ... 'export PATH=...; source ~/.zshrc; which claude'` → "command not found"
- `source ~/.zshrc; echo $PATH` → `"/run/user/0/fnm_multishells/...":"$PATH"` (broken)
Fix:
- Remove `source ~/.bashrc` and `source ~/.zshrc` from ALL launch commands
- ssh -t creates a pseudo-terminal, so bash auto-sources .bashrc for env vars
- Explicit PATH export is all we need for finding the claude binary
- Remove fnm eval snippet from _finalize_claude_install (it poisoned rc files)
- Also: clean up stale ~/.bash_profile, fix cloud-init PATH, move node
install after bun attempt
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Stop writing env vars to ~/.profile and ~/.bash_profile — only write to
.bashrc and .zshrc. The .profile approach caused issues because login
shells source it inconsistently across distros, and creating .bash_profile
makes bash -l skip .profile entirely.
Replace `bash -lc claude` launch commands with explicit PATH export +
source pattern across all cloud providers. This ensures claude is found
regardless of shell initialization quirks.
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two fixes:
1. Swap fallback order from curl → npm → bun to curl → bun → npm.
Bun is faster and typically pre-installed. Use `bun i -g`.
2. Fix "claude: command not found" at launch. The default .bashrc has
a non-interactive guard (`case $- in *i*) ;; *) return;; esac`)
that skips PATH exports when sourced from SSH command strings.
Fix: write env config to ~/.profile (always sourced by login shells)
in addition to .bashrc/.zshrc, and launch with `bash -lc claude`
which starts a login shell that sources ~/.profile.
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
After installing Claude Code (via any method), run `claude install --force`
to set up shell integration, then ensure fnm bootstrap is persisted to both
.bashrc and .zshrc so interactive sessions can find node.
Also simplify all launch commands across 9 clouds: instead of hardcoding
PATH entries that may miss fnm, source the rc files which now contain all
the necessary PATH entries from both inject_env_vars and _finalize_claude_install.
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously offer_github_auth prompted interactively inside inject_env_vars_*,
which runs after the server is already provisioned. This means the user sits
through provisioning before being asked a simple yes/no question.
Split into two phases:
- prompt_github_auth: asks the question early (before create_server)
- offer_github_auth: executes the install later (after server is up),
using the stored answer without re-prompting
Falls back to interactive prompt if prompt_github_auth was never called,
so non-claude scripts and older clouds keep working unchanged.
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All cloud claude.sh scripts had inline curl-only installs with no fallback.
When the curl installer failed (transient outage, rate limit), installation
failed with no recovery. Additionally, fnm-installed Node.js was invisible
to subsequent SSH sessions because each SSH command runs in a non-interactive
shell that doesn't source .bashrc/.zshrc.
Changes:
- Migrate 8 cloud scripts to use shared install_claude_code (curl → npm → bun)
- Move _ensure_node_runtime before npm/bun install attempts (not after)
- Add fnm paths to claude_path so node is discoverable across SSH sessions
- Prefix npm/bun install commands with claude_path for PATH visibility
- Update test assertion to match new install_claude_code behavior
Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
API keys and env vars were only written to .zshrc, so SSH sessions using
bash couldn't find credentials. Also fixes incorrect ~/.claude/local/bin
PATH (claude installs to ~/.local/bin) and syncs interactive_session PATH
with cloud-init PATH across all 9 clouds.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
After an interactive SSH session ends, users are now shown:
- A warning that their server is still running (and may incur charges)
- A link to the cloud provider's dashboard to manage/delete it
- The SSH command to reconnect
This prevents users from unknowingly leaving servers running after
exiting their agent session. Covers all 25 SSH-based cloud providers.
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Unquoted `<< EOF` heredocs in nanoclaw .env file creation cause shell
expansion of the API key value. If an API key contains `$`, backticks,
or `\`, the value is silently corrupted or could trigger command
execution. Replace with `printf '%s'` which safely writes the value
without interpretation.
Also fix unquoted variable expansion in upload_config_file's mv command
and the github-codespaces/openclaw.sh config heredoc.
Fixes 34 scripts across all cloud providers.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- fly/lib/common.sh: Replace 23-line get_server_name() that duplicated
env-var-check, prompt, and validation logic with a one-line call to the
shared get_validated_server_name helper, matching all other cloud providers.
- oracle/lib/common.sh: Break _setup_vcn_networking (48 lines, 3 distinct
responsibilities) into focused helpers:
- _create_internet_gateway: creates the IGW resource
- _add_default_route: configures the route table
- _add_ssh_security_rules: opens SSH port in the security list
The orchestrator _setup_vcn_networking now delegates to these three helpers.
Agent: complexity-hunter
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Add log_install_failed helper to shared/common.sh that provides
structured troubleshooting for agent install failures: possible causes,
SSH debug command (when server IP available), manual install command,
and re-run suggestion. Also improve SSH key registration error message.
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- GCP: capture gcloud stderr on failure, add common issues guidance,
use _log_diagnostic for ensure_gcloud errors
- AWS Lightsail: add common issues for create_server failure,
use _log_diagnostic for ensure_aws_cli errors,
improve instance timeout message with actionable steps
- Cherry Servers: use extract_api_error_message instead of raw response
dump, add common issues for server creation failure
- Oracle Cloud: capture OCI CLI stderr on instance launch failure,
add common issues for VCN, subnet, and instance creation errors
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Hetzner lib: replace all Python JSON parsing with jq. Uses the
/datacenters API as the authoritative source for server type
availability (server_types.available), cross-referenced with
/server_types for specs and pricing. jq is auto-installed if missing.
URLs: update openrouter.ai/lab/spawn → openrouter.ai/labs/spawn
across all READMEs and CLI source.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The get_model_id_interactive function returned MODEL_ID from env vars
without calling validate_model_id, bypassing the allowlist check. Also
migrated 13 legacy scripts from raw safe_read to get_model_id_interactive
which includes validation.
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
~1500 progress messages across 481 files were using log_warn (yellow)
for normal status updates like "Installing...", "Setting up...",
"Creating server...", etc. This made users think something was wrong
when everything was proceeding normally.
Changes:
- Replace log_warn with log_step for all progress/status messages
- Keep log_warn only for actual warnings (errors, remediation hints)
- Remove emoji from 3 sprite completion messages
Agent: ux-engineer
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Add ssh_run_server, ssh_upload_file, ssh_interactive_session, and
ssh_verify_connectivity to shared/common.sh. These four functions
were copy-pasted identically across 21 cloud provider lib files,
differing only in SSH username (root vs ubuntu).
Providers now set SSH_USER and delegate to the shared helpers via
one-line wrappers, reducing each provider's lib by ~20 lines.
Agent: complexity-hunter
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace vulnerable heredoc patterns across 27 continue.sh scripts with
setup_continue_config() helper that uses json_escape() + upload_config_file()
to safely handle API keys containing special characters like quotes or braces.
Also fix _save_token_to_config() in shared/common.sh which had the same
unescaped heredoc vulnerability for local token storage.
Relates to #104
Agent: security-auditor
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Break down _get_subnet_id() (101 lines) and create_server() (87 lines)
into smaller, single-responsibility functions:
- _create_vcn(): VCN creation (19 lines)
- _setup_vcn_networking(): Internet gateway, route table, security list (47 lines)
- _create_subnet(): Subnet creation with AD lookup (25 lines)
- _get_subnet_id(): Now just finds or orchestrates creation (22 lines)
- _get_instance_public_ip(): VNIC lookup and IP extraction (27 lines)
- create_server(): Now delegates IP retrieval (59 lines)
Agent: complexity-hunter
Co-authored-by: A <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
Implement gptme agent script for Oracle Cloud Infrastructure.
Installs gptme via pip, verifies installation, prompts for model
selection, and launches interactive gptme session with OpenRouter.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement Open Interpreter agent deployment on Oracle Cloud
Infrastructure. Installs via pip, configures OpenRouter API
integration via OPENAI_BASE_URL proxy.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement Codex CLI agent deployment on Oracle Cloud Infrastructure.
Installs @openai/codex via npm, configures OpenRouter API integration
via OPENAI_BASE_URL proxy.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement Kilo Code agent deployment on Oracle Cloud Infrastructure.
Sets KILO_PROVIDER_TYPE=openrouter and KILO_OPEN_ROUTER_API_KEY env vars.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Implement Cline agent script for Oracle Cloud Infrastructure.
Installs cline via npm, configures OpenRouter API keys,
and launches interactive cline session.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement NanoClaw agent deployment on Oracle Cloud Infrastructure.
Installs tsx and clones nanoclaw repo, configures .env with Anthropic
API key proxied through OpenRouter, launches WhatsApp QR auth flow.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement Amazon Q CLI agent script for Oracle Cloud Infrastructure.
Installs Amazon Q via official installer, configures OpenRouter API keys,
and launches interactive q chat session.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement OpenClaw agent deployment on Oracle Cloud Infrastructure.
Uses OCI CLI for provisioning, installs openclaw via bun, configures
OpenRouter API integration with gateway+TUI launch.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement Gemini CLI agent script for Oracle Cloud Infrastructure.
Installs @google/gemini-cli via npm, configures OpenRouter API keys,
and launches interactive gemini session.
Agent: gap-filler
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add OCI as a new cloud provider using the OCI CLI for compute instance
provisioning. Includes lib/common.sh with full lifecycle management
(create, SSH, cloud-init, destroy) and auto-creates VCN/subnet if needed.
Implements claude, aider, and goose agent scripts. Supports Always Free
tier shapes (VM.Standard.E2.1.Micro) and flex shapes.
Agent: cloud-scout
Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
