DeployMesh

Portable, provider-neutral deployment intelligence for coding agents.

DeployMesh is an open Agent Skill that makes automatic deployment a required part of application development. It guides an agent from a single GitHub repository to a working production environment on Cloudflare, Railway, Galaxy Cloud, Scalingo, Heroku, or a VPS.

The core promise is simple:

Connect GitHub → choose a target → authorize → deploy → everything required works.

DeployMesh covers infrastructure provisioning, PostgreSQL, queues, Redis, object storage, workers, cron, secrets, environment variables, migrations, domains, HTTPS, health checks, rollback, and automatic redeployment. It does not create separate codebases for different providers.

Why DeployMesh

Most application instructions stop at “the code is ready.” DeployMesh treats deployment as part of the application contract. Agents must detect the required topology, use one repository and one source commit, provision only required resources, connect generated credentials automatically, run migrations, verify every required process, and report blocked or unsupported capabilities honestly.

The recommended portable baseline is React + Vite + TypeScript for the frontend, Hono + TypeScript for the API, PostgreSQL as the canonical relational database, Drizzle for migrations, provider-neutral queue and storage ports, a dedicated worker, and a dedicated cron command. Cloud-specific services remain behind adapters so the application remains portable.

Supported agents

DeployMesh follows the open Agent Skills specification. Its standard SKILL.md package is designed to work with the following agents:

Agent	Project installation	Global installation	Direct use
OpenCode	.opencode/skills/deploymesh/ or .agents/skills/deploymesh/	~/.config/opencode/skills/deploymesh/ or ~/.agents/skills/deploymesh/	Load the deploymesh skill
Claude Code	.claude/skills/deploymesh/	~/.claude/skills/deploymesh/	/deploymesh
Codex	.agents/skills/deploymesh/	~/.agents/skills/deploymesh/	$deploymesh or skill selector
MiniMax Skills	Use Claude Code, Codex, or OpenCode paths documented by MiniMax	Use the corresponding project skill path or --dest	MiniMax-AI/skills documents the supported integrations
Hermes Agent	External skill directory or ~/.hermes/skills/deploymesh/	~/.hermes/skills/deploymesh/	/deploymesh
Cline	.cline/skills/deploymesh/	~/.cline/skills/deploymesh/	Enable or invoke from the Skills panel
Kilo Code	.kilo/skills/deploymesh/ or .agents/skills/deploymesh/	~/.kilo/skills/deploymesh/	Reload skills, then invoke deploymesh

The repository uses the portable Agent Skills frontmatter required by the standard and plain Markdown instructions. OpenCode, Claude Code, Codex, MiniMax Skills, Hermes, Cline, and Kilo Code can use the package through their documented skill paths. Any other Agent Skills-compatible agent can use the generic --dest installer mode. See Agent compatibility for current links and exact paths.

Platform skill suite

DeployMesh includes one core skill and separate platform adapters. Always load deploymesh first, then load the adapter for the selected target. The adapter does not replace the core contract; it adds provider-specific deployment, authorization, scheduler, service, and verification rules.

Skill	Target	Official documentation	Official external skill
cloudflare-deploy	Cloudflare	Docs for agents, Workers Builds, API	cloudflare/skills
scalingo-deploy	Scalingo	Scalingo API, Deployments, Scheduler	No official provider skill confirmed; DeployMesh adapter included
railway-deploy	Railway	GitHub Autodeploys, SaaS backend, Public API	Railway Agent Skills, railwayapp/railway-skills
heroku-deploy	Heroku	Platform API, GitHub Integration, Scheduler	No official provider skill confirmed; DeployMesh adapter included
galaxy-deploy	Galaxy Cloud	Galaxy docs, Node.js, PostgreSQL	No official provider skill confirmed; capability discovery required

See the complete platform skill guide. If an official provider skill exists, prefer it for the provider's product-specific behavior and use the DeployMesh adapter for cross-platform orchestration, GitHub source control, migrations, health checks, rollback, and final status reporting.

Install everywhere

The simplest installation is to clone the repository and run the included installer:

git clone https://github.com/prohunter00017/deploymesh.git ~/.deploymesh
bash ~/.deploymesh/scripts/install.sh --all

The installer creates symlinks into the supported global skill directories and never overwrites an existing skill without an explicit --force flag. Restart the agent, or use its skill reload command when available.

Install every platform adapter for every supported agent:

bash ~/.deploymesh/scripts/install.sh --all --all-platforms

Install through MiniMax's documented Claude Code, Codex, and OpenCode environments:

bash ~/.deploymesh/scripts/install.sh --minimax
bash ~/.deploymesh/scripts/install.sh --minimax --all-platforms

Install into any other Agent Skills-compatible agent by passing its configured skill root:

bash ~/.deploymesh/scripts/install.sh --dest ~/.my-agent/skills
bash ~/.deploymesh/scripts/install.sh --dest ~/.my-agent/skills --all-platforms

The generic mode creates DIR/deploymesh and, when requested, DIR/<platform>-deploy symlinks. It does not assume or invent a vendor-specific directory.

Install only the adapters you need for one agent:

bash ~/.deploymesh/scripts/install.sh --claude-code --railway --heroku
bash ~/.deploymesh/scripts/install.sh --codex --cloudflare
bash ~/.deploymesh/scripts/install.sh --opencode --scalingo

Platform adapters are named cloudflare-deploy, scalingo-deploy, railway-deploy, heroku-deploy, and galaxy-deploy. Load only the selected adapter to keep the agent's context focused.

Install only one agent when preferred:

bash ~/.deploymesh/scripts/install.sh --claude-code
bash ~/.deploymesh/scripts/install.sh --codex
bash ~/.deploymesh/scripts/install.sh --opencode
bash ~/.deploymesh/scripts/install.sh --hermes
bash ~/.deploymesh/scripts/install.sh --cline
bash ~/.deploymesh/scripts/install.sh --kilo

For MiniMax, use --minimax or the generic --dest mode. For a project-local installation on any agent, clone or copy this repository into the corresponding project skill root. Project-local installation is recommended when a team wants the deployment standard versioned with the application.

Use it

Ask the agent to use DeployMesh before building or deploying an application:

Use the deploymesh skill before starting this application. Build it as one portable GitHub repository and make deployment automatic for Railway and a VPS.

For Claude Code or Hermes Agent, invoke it explicitly:

/deploymesh design the application and its automatic deployment topology

For Codex, use the skill selector or explicitly reference $deploymesh. For OpenCode, Cline, and Kilo Code, ask the agent to load the deploymesh skill or enable it in the agent's skill settings.

Repository structure

SKILL.md                         Core agent instructions
references/
  architecture-and-stack.md     Portable stack and repository layout
  deployment-contract.md        Manifest, state machine, secrets, verification
  provider-adapters.md           Cloud and VPS adapter rules
skills/
  cloudflare-deploy/SKILL.md     Cloudflare adapter and official skills link
  scalingo-deploy/SKILL.md       Scalingo API and scheduler adapter
  railway-deploy/SKILL.md        Railway GitHub and service adapter
  heroku-deploy/SKILL.md         Heroku API and release adapter
  galaxy-deploy/SKILL.md         Galaxy capability-discovery adapter
research-notes.md                Source-backed capability notes
templates/
  deployment.manifest.yaml       Provider-neutral deployment manifest template
docs/
  agent-compatibility.md         Installation paths and official agent links
  platform-skills.md             Platform catalog and official references
platform-skill-research.md       Source-backed provider skill research
scripts/
  install.sh                     Safe global installer
  validate.sh                    Local validation wrapper
.github/workflows/
  validate.yml                   Pull-request and push validation

Supported deployment targets

Target	Native strategy
Cloudflare	Workers/Pages, Workers Builds, R2, Queues, Cron Triggers, secrets, routes, and API/Wrangler adapters
Railway	GitHub-linked services, API/web, Postgres, Redis, worker, cron service, reference variables, and CI gates
Galaxy Cloud	GitHub-linked app deployment, documented runtime/database capabilities, and runtime capability checks
Scalingo	Scalingo API, managed add-ons, environment variables, deployments, domains, and Scheduler with limit validation
Heroku	Platform API/OAuth, GitHub integration, Postgres/Redis add-ons, config vars, process formation, release phase, and Scheduler
VPS	Automated SSH bootstrap, Docker Compose, PostgreSQL, Redis, worker, cron, Caddy, HTTPS, migrations, health checks, and rollback

The skill deliberately distinguishes a provider feature from a portable guarantee. For example, Scalingo and Heroku document scheduler limitations, while the reviewed Galaxy documentation does not establish a public provisioning API for every resource. DeployMesh therefore uses capability discovery and reports BLOCKED, UNSUPPORTED, or DEGRADED instead of inventing support.

Quality and security

DeployMesh requires idempotent reconciliation, immutable commit tracking, migration locking, health checks, redacted logs, least-privilege authorization, and rollback retention. It forbids committing or printing secrets. It also forbids destructive production changes without explicit confirmation.

Run local validation with:

bash scripts/validate.sh

License

DeployMesh is released under the MIT License.

References