About Hosting EveAgents

EveAgents is an open-source directory of agents built for the Eve framework. Each agent includes its instructions, source files, configuration, requirements, and optional integration-specific files.

The Railway template downloads the selected agent during deployment, validates every file, applies Railway-compatible runtime adapters, and starts the agent on Railway’s assigned port.

To get started:

Choose an agent from EveAgents.
Create a key from the EveAgents API Keys page.
Paste it into EVEAGENTS_API_KEY when Railway asks.
Choose OpenAI, Claude, or Gemini and add the corresponding provider API key.
Add any credentials required by the selected integration.
Deploy the service.

EVEAGENTS_API_KEY is the only EveAgents-issued credential required by the template.

Why Deploy EveAgents on Railway?

Railway provides a convenient environment for running persistent Eve agents without manually configuring a server.

The template includes:

Automated agent download and file-integrity validation
Direct OpenAI, Claude, and Gemini model support
Railway-compatible Eve runtime configuration
Public HTTP networking
Health monitoring through /eve/v1/health
Persistent workflow storage
Automatically generated route authentication
Support for standalone agents and compatible integration variants
Rebuilds and redeployments from one maintained template

Common Use Cases

Deploy productivity agents for meetings, planning, and task management
Run engineering agents for incident response, bug triage, and releases
Connect agents to Discord, Telegram, Microsoft Teams, and other channels
Build research, knowledge-management, support, and analytics workflows
Combine a reusable base agent with integration-specific functionality
Run internal agents behind password-protected Eve routes

Dependencies for EveAgents Hosting

The following accounts and credentials may be required:

A Railway account
An EveAgents account
An EveAgents API key
An API key from one supported AI provider:
OpenAI
Claude
Gemini
Any credentials required by the selected integration

Integration-specific requirements are listed on the corresponding agent integration page in the Eve integrations directory.

Deployment Dependencies

How Distribution Works

EVE_AGENT_SLUG selects a published base agent.
EVE_INTEGRATION_SLUG optionally selects an assigned integration variant.
The build uses EVEAGENTS_API_KEY to download the selected agent.
Integration-specific files are applied over the reusable base files.
Every file path, size, and SHA-256 digest is validated.
Required environment variables are checked before the agent is built.
The template selects the direct model adapter identified by EVE_MODEL.
The template adds the Railway-compatible Eve web channel and sandbox adapters.
Eve builds and starts the selected agent on Railway’s assigned PORT.

Each EveAgents API key allows up to 20 agent downloads per hour. Keys can be revoked from the API Keys page.

Railway Variables

EVE_AGENT_SLUG — Required. Selects the published agent to deploy (i.e. incident-response-commander). Browse Eve agents.
EVE_INTEGRATION_SLUG — Selects an optional integration variant; leave it empty for the standalone agent (i.e. slack or discord). Browse integrations.
EVEAGENTS_REGISTRY_URL — Defines the registry endpoint used to download the selected agent (i.e. https://www.eveagents.dev/api/registry/v1). This value is preconfigured and should not be changed.
EVEAGENTS_API_KEY — Required. Authorizes the template to download an agent from EveAgents (i.e. eva_live_...). Create one on the EveAgents API Keys page.
EVE_MODEL — Required. Selects the AI provider and model used by the agent (i.e. openai/gpt-5.4-mini, anthropic/claude-sonnet-4-6, or google/gemini-3.6-flash).
EVE_PROVIDER_API_KEY — Required. The direct API key for the provider selected in EVE_MODEL (i.e. use an OpenAI key when EVE_MODEL starts with openai/).
ROUTE_AUTH_BASIC_USER — Defines the username protecting the deployed agent’s public Eve routes (i.e. eve). This value is preconfigured.
ROUTE_AUTH_BASIC_PASSWORD — Defines the password protecting the deployed agent’s public Eve routes (i.e. a Railway-generated 32-character secret). It is preconfigured using ${{ secret(32) }}.

Railway Service Settings

Source: https://github.com/bergside/eve-railway-template
Config file: /railway.json
Public networking: HTTP enabled
Healthcheck: /eve/v1/health
Persistent volume: /app/.eve/.workflow-data

Current Integration Compatibility

Standalone agents and direct-credential channels can run entirely on Railway.

Some Slack and MCP/OpenAPI connection variants currently rely on Vercel Connect. These variants require a separately configured Connect authorization flow because Railway does not provide Vercel deployment identity.

Review the requirements on the selected integration page before deploying.

License

MIT