Reference¶
This section provides a detailed reference for the a5c project, covering public API interfaces, configuration options, and available CLI command patterns.
API Reference¶
Public API definitions, endpoint specifications, and data models are maintained in the a5c-ai/spec repository. For comprehensive interface details, refer to:
https://github.com/a5c-ai/specs
Configuration Options¶
a5c uses a YAML configuration file at the root of your repository named .a5c/config.yml. The primary top-level keys and their defaults are shown below:
defaults:
# Command template for invoking AI agents (adjust as needed for your environment).
cli_command: "<your model execution command template>"
remote_agents:
# Enable or disable fetching agents from the public registry
enabled: true
# Cache timeout for registry responses (minutes)
cache_timeout: 120
# Number of retry attempts when fetching agents
retry_attempts: 5
# Delay between retries (milliseconds)
retry_delay: 2000
sources:
individual:
- uri: "<agent registry URI>"
alias: "<agent-alias>"
Option |
Type |
Default |
Description |
|---|---|---|---|
|
string |
See snippet above |
Shell command template used to invoke the AI model/agent. |
|
boolean |
|
Toggle remote registry agent support. |
|
integer |
|
Minutes to cache registry responses. |
|
integer |
|
Number of retry attempts fetching remote definitions. |
|
integer |
|
Delay in milliseconds between retries. |
|
list |
N/A |
Array of agent definitions with |
For additional configuration examples and advanced settings, consult the full .a5c/config.yml in your repository.
CLI Commands¶
a5c does not currently provide a dedicated command-line interface. Instead, agent invocation is controlled via the cli_command template in .a5c/config.yml. This allows you to integrate any shell-based AI client (e.g., codex exec, claude, or openai) as needed.
Selecting the AI client¶
You can configure which AI client to use by setting the A5C_CLI_TOOL environment variable in your GitHub Actions variables. Valid values are:
claudecodexazure_codexgemini
Set the variable at the repository or organization level under Settings → Secrets and variables → Actions → Variables.