# Kenn Software: local-first agent tooling suite

Kenn Software ([kenn.io](https://kenn.io), tagline "development and knowledge systems for the agentic era") ships a small family of tools that all look the same from a distance: a single Go binary, local-first, SQLite under the hood, no account or telemetry, a CLI shaped for agents and a TUI shaped for the human watching them. Five of them, covering different points of the agent-assisted-development loop.

## The five tools

- [[kata]] — issue tracker. A local SQLite daemon with an agent CLI and a human TUI; issue state lives *outside* the repo (only a `.kata.toml` is committed). The durable task ledger an agent records against. *(210★, early preview — [[toolbox/watchlist]])*
- [[middleman]] — PR/issue dashboard. Syncs GitHub/GitLab/Forgejo/Gitea data into local SQLite and serves a Svelte UI; unified activity timeline, inline diffs, a local kanban, merge/approve from the UI. For triaging *other people's* PRs. *(110★, no release yet)*
- [[agentsview]] — session analytics. Full-text search, cost tracking (a 100×-faster `ccusage` replacement), and stats across 50+ agents. The volume-and-cost ledger. *(~4.5k★ — the mature one)*
- [[roborev]] — continuous code review. A post-commit hook reviews every commit in the background and can auto-fix in isolated worktrees. The output-quality ledger. *(1.2k★, v0.55)*
- [[msgvault]] — email/chat archival. Gmail/IMAP/MBOX/.emlx into SQLite + DuckDB; offline FTS and vector search, MCP server. The personal-data member, not an agent-workflow tool. *(1.8k★, v0.14 alpha — [[toolbox/watchlist]])*

## The shared design DNA

The reason these belong on one page is that they make the same bets, and the bets are the interesting part:

- **Single binary, local-first.** Everything runs on your machine. Servers bind to `127.0.0.1` by default. No hosted service, no accounts, no telemetry beyond an optional update check.
- **SQLite as the substrate**, often with a second engine for what SQLite is bad at — DuckDB + Parquet in msgvault for OLAP aggregates, FTS5 across agentsview/msgvault for search. middleman uses the pure-Go `modernc.org/sqlite` driver to avoid CGO; agentsview and msgvault take the CGO/static-linking cost where they need FTS5 or DuckDB.
- **CLI for agents, TUI for humans.** Stable commands with JSON output and predictable failure modes on one side; vim-keyed interactive browsing on the other. kata states this split explicitly; the others follow it.
- **Speed via pre-indexing.** agentsview's headline claim — 100× faster than re-parsing session JSONL on every run — is the same instinct behind msgvault's DuckDB cache: index once, query in milliseconds.
- **Optional Postgres for teams.** Both agentsview and roborev let you push local data to a shared Postgres for team dashboards, while staying single-user-local by default.
- **Go 1.25/1.26, MIT.** Same toolchain and license across the board.

## Where they sit relative to other wiki tools

The suite overlaps deliberately with tools already tracked here, and the contrasts are sharper than the similarities:

- **kata vs. [[beads]]** — both give agents a structured task graph instead of a markdown TODO, but opposite storage philosophies. Beads commits a Dolt database *into* the repo for branch-merged issues; kata keeps SQLite *outside* the repo for a clean footprint. Pick by whether you want issue history versioned with code.
- **middleman vs. [[forge]]** — same set of forges (GitHub/GitLab/Gitea/Forgejo). forge is the scripting/agent CLI bridge; middleman is the human review surface.
- **roborev vs. [[charlie-daemons]] / [[toolbox/re_gent]]** — roborev is the commit-triggered, review-shaped case of "agent runs in the background"; re_gent attacks the same "accountability for generated code" goal through per-prompt provenance instead of review.
- **agentsview vs. [[ctx]]** — ctx organizes and binds agent conversations; agentsview analyzes and prices them.
- **msgvault vs. [[wacli]]** — same "own your message history offline" instinct, different source (email vs. WhatsApp).

The through-line worth keeping: this is one vendor betting that the tooling around coding agents should be local single binaries you own, not SaaS dashboards — and shipping enough of them (issues, PRs, cost, review, personal data) to make the bet legible.