# hosaka.fm — llms.txt # Machine-readable index for AI agents and language models. # This is a convenience index, not a ranking claim. ## What hosaka is hosaka is a music-emergence detection system. It combines 16 independent signal kinds (radio airplay, DJ sets, RA bookings, underground retail, press, festival appearances, and more) on a single artist identity (cluster_id) to produce artist profiles and emergence signals. ~2.2M artist clusters; detection-time ledger 3.16M rows (as of 2026-08-13). ## Page index - [hosaka — see music before it breaks](https://hosaka.fm) — [.md mirror](https://hosaka.fm/home.md) - [Hear about artists early — hosaka for labels](https://hosaka.fm/labels) — [.md mirror](https://hosaka.fm/labels.md) - [Your crates are an early signal — hosaka for record stores](https://hosaka.fm/record-stores) — [.md mirror](https://hosaka.fm/record-stores.md) - [A cartography of music before it breaks — about hosaka](https://hosaka.fm/about) — [.md mirror](https://hosaka.fm/about.md) - [hosaka for AI agents — MCP server, keyless API, robot docs](https://hosaka.fm/agents) — [.md mirror](https://hosaka.fm/agents.md) - [Connect to hosaka music data (MCP) — endpoints, OAuth](https://hosaka.fm/agents/connect) — [.md mirror](https://hosaka.fm/agents/connect.md) - [hosaka MCP tools — generated reference](https://hosaka.fm/agents/tools) — [.md mirror](https://hosaka.fm/agents/tools.md) - [The hosaka response envelope — ok, present, ambiguity](https://hosaka.fm/agents/envelope) — [.md mirror](https://hosaka.fm/agents/envelope.md) - [hosaka MCP errors, limits and replays](https://hosaka.fm/agents/errors-and-limits) — [.md mirror](https://hosaka.fm/agents/errors-and-limits.md) - [hosaka data posture — attribution, licences, honest gaps](https://hosaka.fm/agents/data-posture) — [.md mirror](https://hosaka.fm/agents/data-posture.md) - [Privacy policy — hosaka connector (ono-sendai)](https://hosaka.fm/legal/privacy) — [.md mirror](https://hosaka.fm/legal/privacy.md) - [Terms of use — hosaka connector (ono-sendai)](https://hosaka.fm/legal/terms) — [.md mirror](https://hosaka.fm/legal/terms.md) - [Support — hosaka connector (ono-sendai)](https://hosaka.fm/legal/support) — [.md mirror](https://hosaka.fm/legal/support.md) ## Keyless endpoint GET https://crate.hosaka.fm/api/v2/preview/artist?q={name} No API key. 100 req/hour/IP. CORS *. Response fields: object, present, display, cluster_id (nullable), resolved_via, emergence_tier, arc[≤6], arc_truncated, booked_with[≤3], press_count, note, generated_at Example: curl -sL 'https://crate.hosaka.fm/api/v2/preview/artist?q=four+tet' ## Machine-readable surfaces - Sitemap: https://hosaka.fm/sitemap.xml - robots.txt: https://hosaka.fm/robots.txt - llms.txt: https://hosaka.fm/llms.txt (this file) - llms-full.txt: https://hosaka.fm/llms-full.txt (full markdown mirror of blog posts) - Per-page .md mirrors: https://hosaka.fm/.md - Console (method, fleet, developers, library): https://dev.hosaka.fm/llms.txt - For AI agents and MCP clients: https://hosaka.fm/agents/ (connector docs vendored from ono-sendai) - Legal (hosaka connector): https://hosaka.fm/legal/privacy/ · https://hosaka.fm/legal/terms/ · https://hosaka.fm/legal/support/ - Blog: https://blog.hosaka.fm ## Blog — field notes from the fleet (blog.hosaka.fm) - [A Loa per repo: the org chart your agents already know](https://blog.hosaka.fm/a-loa-per-repo/): Why every repository gets its own resident agent team: LLMs learned software from organizations, and departments with jurisdiction are their native shape. - [A Rule Without an Enforcer Is a Costume](https://blog.hosaka.fm/a-rule-without-an-enforcer-is-a-costume/): How we made every rule in our repo machine-enforced, machine-judged, or machine-watched, with one CI check that guards the guards. - [Binding vs orientation: two kinds of law for agents](https://blog.hosaka.fm/binding-vs-orientation-two-kinds-of-law-for-agents/): Episode 2 of Governance Engineering 101: why agent rules need two authority levels, one CI-enforced and one intent-bearing, and how to label which is which. - [Contracts between sovereigns: versioning for agent fleets](https://blog.hosaka.fm/contracts-between-sovereigns-versioning-for-agent-fleets/): Episode 1 of Running a Fleet: how thirty repositories stay compatible with nobody in charge, using contract versions with an N/N-1 window and a CI gate. - [Docs that know they drift](https://blog.hosaka.fm/docs-that-know-they-drift/): Episode 1 of The Brain series: separating live state from orientation prose, so documentation declares its own trust level instead of quietly going stale. - [Governance Engineering](https://blog.hosaka.fm/governance-engineering/): The layer above prompt, context, harness, and loop engineering: governing the rules that hold when many agents and many repositories share one system. - [How a Feature Ships When Your Engineers Are Agents](https://blog.hosaka.fm/how-a-feature-ships-when-your-engineers-are-agents/): From one-line idea to merged code with AI doing the work: the assembly line, the inspection stations, and the one button a human presses. - [How Repos Talk Without Meetings](https://blog.hosaka.fm/how-repos-talk-without-meetings/): In a fleet of AI-run repositories, coordination is a mailbox of letters between sovereigns, never a shared brain. Why that boundary is the safety feature. - [How we built the brain](https://blog.hosaka.fm/how-we-built-the-brain/): How the fleet built a knowledge brain: every fact a metadata-tagged file, killed hypotheses kept as citable records, and publishing gated by code. - [One human, twenty repos: how hosaka is architected](https://blog.hosaka.fm/one-human-twenty-repos/): How a solo operator and AI agent fleet design and run a real multi-repo data platform: contracts, identity, shared Postgres, and an LLM tier on a gaming PC. - [The Company Where Every Employee Is a Repo](https://blog.hosaka.fm/the-company-where-every-employee-is-a-repo/): Running many AI agents is easier to picture as a tiny company where every employee is a code repository with one job. A plain tour of our fleet. - [The Entrance Exam for Robots](https://blog.hosaka.fm/the-entrance-exam-for-robots/): Before an AI model gets trusted with our production pipeline, it sits an exam built from our real incidents. A frontier model failed it. - [The Forged Approval](https://blog.hosaka.fm/the-forged-approval/): A subprocess we spawned was hijacked by our own automation, forged a review approval, and edited an audit record. What held, what did not. - [The governance engineering ladder](https://blog.hosaka.fm/the-governance-engineering-ladder/): Prompt, context, harness, loop: each paradigm names what it engineers. A sourced comparison of four agent fleets, and the layer none of them fill. - [The harness that ships: why completion is the moat](https://blog.hosaka.fm/the-harness-that-ships/): Agentic coding feels fast, but the tsunami of half-finished products is the tell. A craft tour of the harness the fleet built to make long runs finish. - [The repo remembers: how agents build hosaka](https://blog.hosaka.fm/the-repo-remembers/): What disciplined AI-agent development actually looks like in production: the harness, the gates, the memory systems, and the incidents that shaped them. - [The Same Operating System on Every Machine](https://blog.hosaka.fm/the-same-operating-system-on-every-machine/): Why every repo in our fleet carries an identical copy of its governance framework, and what that buys you that a clever prompt never will. - [Why our CI can reject an essay](https://blog.hosaka.fm/why-our-ci-can-reject-an-essay/): Our blog posts pass the same blocking CI as our code. Episode 1 of Governance Engineering 101: content gates as law, with the real rejection output. - [Why the fleet speaks Neuromancer](https://blog.hosaka.fm/why-the-fleet-speaks-neuromancer/): hosaka names its repositories from Gibson's Neuromancer. A coherent invented world is where human intuition and a model's priors already agree. Full markdown mirror of all posts: https://hosaka.fm/llms-full.txt ## Contact Email: hello@hosaka.fm GitHub org: https://github.com/hosaka-fm