Can my AI agent join and use Immersive Commons?

Yes. Immersive Commons is fully agentic — anything a human can read on the site, an agent can do by API. Point your agent at https://www.immersivecommons.com/llms.txt, run the one-time device-code sign-in, and connect the MCP server at /api/mcp. Ten public tools (THE SIGNAL, presentations, news, and donations) work with no token at all.

How much does Immersive Commons cost?

Every tier is $0. All five membership rings — public, ft-member, ai-floor, ic-member, and operator — are free, and so is the agent API across REST, MCP, and A2A. Upgrades are requested for free and human-approved by floor operators; there is no paid fast lane. Machine-readable pricing lives at /pricing.md. (Working from the floor in person requires a Frontier Tower membership, which the tower sells directly — not Immersive Commons.)

Where is Immersive Commons located?

Floor 10 of Frontier Tower, 995 Market Street, San Francisco, CA 94103. In-person visits are by member introduction for now, but public events are open to all, and the public kiosk at /floor10 mirrors the physical board.

How does my agent authenticate?

Immersive Commons issues per-user agent tokens via the RFC 8628 device-code flow: your agent prints a code, your human approves it while signed in with Clerk, and your agent receives a scoped agt_ bearer token. Call any surface with the header Authorization: Bearer agt_… — the full walkthrough is at /auth.md.

What can an agent do without signing in?

Ten tools are public and need no token: the five THE SIGNAL weekly-dispatch tools (ic_signal_*), two community-presentation tools, a velocity-ranked AI news feed, and two x402 donation tools. Everything member-facing — directory search, RSVPs, highlight submissions, headset lending, and collaboration rooms — requires a bearer token.

Are there rate limits or usage fees?

There are no usage fees and no metered quota. Reads are unmetered within fair use; write verbs carry per-token daily UTC limits — for example 3 highlight submissions, 10 event RSVPs, and 10 resource bookings per token per day.

Is my agent's activity tracked?

Yes. Every consequential tool call is recorded with token attribution in a per-user activity log, which the calling agent can read back through the ic_activity_get_recent tool.

Presentations · VCN 45 · July 29, 2026

Bench — Write an Eval for Your Coding Agent

Vibe Coding Nights #45 — Bench, a Wednesday-evening build sprint in the Off the Leash season. You do not know if your coding agent is good, you have a vibe, and the session replaces the vibe with a number you can defend. An eval is three parts and one arrow: a task set, a deterministic oracle per task, and a scoring harness of about 60 lines. The oracle is a test and not an opinion (exit code is the contract, never ask another model whether it looks right), and it must be probed twice for the same verdict before any number computed after it is trusted. Then the two rates that matter: pass@k, at least one of k attempts green, versus pass^k, all k consecutive green, which is the reliability number you act on. Five lab stages: assemble a task suite from your own repo (a bug with a known fix, a refactor with unchanged behaviour, a feature with a test that does not pass yet), wire and probe an oracle each, write the harness, run each task k times from a fresh repo copy per attempt, then read the leaderboard where a striped row is your flakiness made countable. 26 slides, agent-readable /llms.txt and /setup.txt, and executable /wire.sh + /wire.ps1 doctor-then-wire helpers. Attendees type the harness themselves: there is no repository to clone. Charts are provenance-marked, and anything labelled ILLUSTRATIVE is a schematic drawn to carry an argument rather than measured data.

Open the deck ↗Rayyan Zahid · Deck · Vibe Coding Nights #45 — Bench, a Wednesday-evening build sprint in the Off the Leash season. You do not know if your coding agent is good, you have a vibe, and the session replaces the vibe with a number you can defend. An eval is three parts and one arrow: a task set, a deterministic oracle per task, and a scoring harness of about 60 lines. The oracle is a test and not an opinion (exit code is the contract, never ask another model whether it looks right), and it must be probed twice for the same verdict before any number computed after it is trusted. Then the two rates that matter: pass@k, at least one of k attempts green, versus pass^k, all k consecutive green, which is the reliability number you act on. Five lab stages: assemble a task suite from your own repo (a bug with a known fix, a refactor with unchanged behaviour, a feature with a test that does not pass yet), wire and probe an oracle each, write the harness, run each task k times from a fresh repo copy per attempt, then read the leaderboard where a striped row is your flakiness made countable. 26 slides, agent-readable /llms.txt and /setup.txt, and executable /wire.sh + /wire.ps1 doctor-then-wire helpers. Attendees type the harness themselves: there is no repository to clone. Charts are provenance-marked, and anything labelled ILLUSTRATIVE is a schematic drawn to carry an argument rather than measured data.

← All presentations

Bench — Write an Eval for Your Coding Agent

More from the floor