Product Think¶

The agent-as-a-product perspective: how it feels to use, what it costs to run, how predictable it is in the wild. Easy to under-invest in because it's not a research problem — but it's what determines whether anyone keeps the system around.

How reyn handles it¶

CLI affordances¶

reyn's CLI is structured as small, composable subcommands rather than one monolithic entrypoint — each owning exactly its own subsystem's operator surface (agent / topology / memory / permissions / events / mcp / config / …), sharing the same reyn.yaml and .reyn/ state directory rather than a shared mega-command. See feature-map.md's CLI section for the full, current command inventory rather than a duplicated list here.

Live legibility: the inline CUI's audit chips¶

The inline CUI's status-chip bar (Agents / Cost / Model / Tools / MCP / Skills / Hooks / Pipes / Cron / Tasks) surfaces the same operator-visible state the P6 audit-event log records, live and inline rather than only available via after-the-fact replay — this is the dual-facet companion to the Observability lens's reading of the same chips (see observability.md).

Cost reporting and reduction — distinct from bounding¶

Two things that look similar but are lens-distinct:

Cost reporting (this lens): /cost gives a quick token + USD summary for the current agent; /budget gives a full breakdown. cost_warn is a pre-selection warning to the operator when the resolved model's cost-per-1M-tokens exceeds a threshold, de-duped once per model per session — legibility and predictability, nothing more.
Cost bounding (the cross-cutting band's cost/budget member, not this lens): hard per-agent / daily / monthly token+USD caps that refuse further spend once exceeded. Don't cite the bounding caps as a Product Think exemplar — they're the band's job, not this lens's (see CLAUDE.md's Constitution section for the full band↔lens distinction).
Cost reduction is this lens's other facet: present routes bulk data to the surface at ~0 output tokens instead of reproducing it as LLM output — a genuine token-cost reduction, not just a reporting mechanism.

Predictable UX¶

output_language. One config key controls the language of user-facing output. No per-agent localization code.
reyn events. When a run does something unexpected, the artifact-of-record is one CLI call away.
State is on disk. .reyn/ holds events, chats, approvals, memory. Nothing important is in process memory only.
On-limit modes. interactive / auto_extend / unattended give the operator predictable, config-selectable control over every loop/timeout/budget checkpoint uniformly — see reliability-engineering.md.
/agents view. Lists running agents/sessions and lets you attach — operator legibility into orchestrated work, spanning skill runs and delegated peers.

Where it's still thin¶

No cost dashboard or trend view. Per-run cost is shown (/cost, /budget); aggregating across runs is the operator's own job (the data is structured enough to feed into other tools).
Onboarding has rough edges. reyn init scaffolds config, but the getting-started guide is the actual orientation — a single integrated one-command quickstart doesn't exist.

These are addressable without changing the OS — they're product polish on top of an already-stable runtime.