Extending pertmux

pertmux is designed to be extensible. You can add new coding agents and forge integrations by implementing the appropriate traits.

Adding a coding agent

Implement the CodingAgent trait in src/coding_agent/:

pub trait CodingAgent {
    fn name(&self) -> &str;
    fn process_name(&self) -> &str;
    fn query_status(&self, pane: &AgentPane) -> PaneStatus;
    fn send_prompt(&self, pane_pid: u32, session_id: &str, prompt: &str) -> anyhow::Result<String>;
    fn enrich_pane(&self, _pane: &mut AgentPane) {}
    fn fetch_session_detail(&self, _session_id: &str) -> Option<SessionDetail> { None }
}

• name(): Human-readable name for the agent (e.g. "opencode", "claude-code").
• process_name(): The process name to detect in tmux panes (e.g. "opencode", "claude").
• query_status(): Takes the pane info and returns a PaneStatus enum (Busy, Idle, Retry, Unknown).
• send_prompt(): Delivers a prompt to the agent. The implementation determines the delivery mechanism — opencode uses its HTTP API, Claude Code uses tmux send-keys.
• enrich_pane(): Populates pane metadata (session title, model, tokens, etc.) from the agent’s data source.
• fetch_session_detail(): Returns detailed session info for the detail panel.

Each agent is responsible for its own data source — opencode uses HTTP API + SQLite, Claude Code reads JSONL transcript files.

Register your agent in agents_from_config() in src/coding_agent/mod.rs.

Adding a forge

Implement the ForgeClient trait in src/forge_clients/:

#[async_trait(?Send)]
pub trait ForgeClient {
    async fn fetch_mrs(&self) -> anyhow::Result<Vec<MergeRequestSummary>>;
    async fn fetch_mr_detail(&self, iid: u64) -> anyhow::Result<MergeRequestDetail>;
    async fn fetch_ci_jobs(&self, mr: &MergeRequestDetail) -> anyhow::Result<Vec<PipelineJob>>;
    async fn fetch_notes(&self, iid: u64) -> anyhow::Result<Vec<MergeRequestNote>>;
    async fn fetch_discussions(&self, iid: u64) -> anyhow::Result<Vec<MergeRequestNote>>;
    async fn fetch_user_mrs(&self) -> anyhow::Result<Vec<UserMrSummary>>;
}

• fetch_mrs(): Fetch open MRs/PRs for the configured project.
• fetch_mr_detail(): Fetch full detail for a single MR (branch info, merge status, etc.).
• fetch_ci_jobs(): Fetch CI/pipeline jobs for a given MR.
• fetch_notes(): Fetch inline notes/comments for a MR.
• fetch_discussions(): Fetch threaded discussions for a MR.
• fetch_user_mrs(): Fetch all open MRs/PRs authored by the authenticated user across the entire forge (not scoped to a project). Used to power the MR Overview popup.

Each forge handles its own:

• API authentication
• Response parsing
• State normalization (e.g., GitHub "open" becomes "opened")