Click a block to open it; in a block, the small map at the top takes you to any other block. Click a check, a sample, a gate on the trust boundary, a stop or a return to see it in the inspector. Play steps through the scenario. Red text is the LLM's proposal, in its own words.
A gate on the trust boundary shows its first check and how many more: R-01 +5 is R-01 and five more checks made at that crossing. The letters say where a check is made: K containment, X treasury engine, R books check, H Consistency check, MO Policy and figures check, G2 Narrative check, A approval, D Dealer, BO back office, T the books. The tag beside a check says what it decides: return sends the plan back to the AI with reasons; block, escalate, warn, flag and the rest are in Templates › Catalogues.
The second line, under the tabs, holds each tab's own controls: the scenario and Visibility on the Map; Visibility and Build or buy on AI Industry; the slides of the Landscape; the switches on the Architecture; and the sections of Why MERLIN, Build or buy, Platforms and Deployment, as tabs of their own. The tabs on the first line never move.
Visibility, on the second line, sets how much the map shows. Clear is the blocks, the forward flow, the Analyst-run loop and the trust boundary; five options, each a toggle, add one kind of detail each (descriptions, exceptions, people, calls to the books, checks), in any combination; All is everything. When the scenario plays an event on a level you have hidden, that one element shows for the event.
The Why MERLIN tab is the value proposition: a treasury proving ground, what it does, how a company uses it, the playbook it takes away, and the open points that must be true before the claim holds.
The Treasury Landscape tab shows where MERLIN sits in treasury, as five slides; ‹ and › on the second line, or the arrow keys, move between them. Slides 1 and 2 are treasury operations standing alone, with nothing of MERLIN in them: the four layers and their blocks by name, then each block with what it covers. Slide 3 is all of treasury with the daily operation MERLIN covers as the main line; slide 4 the daily operation alone, MERLIN's scope; slide 5 that with MERLIN's blocks laid over it. A MERLIN block opens its close-up, and the Map shows the same blocks in motion.
The Architecture tab shows MERLIN as built: the client, the server with its three agents that each hold one key to the books, the LLM layer beside it, and MERLIN Tools, which is the books and issues the keys. Switch on the interfaces, the keys, or the Level 1 or Level 2 framework blocks; click a box for its code, checks and blocks.
The Platforms tab sets out how agent platforms are built (frameworks such as CrewAI, vendor SDKs, cloud agent runtimes such as AWS AgentCore, managed agents, and app builders) and maps each onto MERLIN: an overview with the two tables, then one tab per family on the second line.
The Deployment tab is the same question from an organisation's side: which parts of an AI system it can run itself and which a cloud provider, a model vendor or a tool vendor can run for it. Its sections are tabs on the second line: the map, the parts, tooling, models and data, sandboxes, integrators, MERLIN placed four ways, and three worked examples. On Examples, a third set of tabs beside them shows All three or one alone: CrewAI, Emergent or Claude Code.
The AI Industry tab shows AI as practised today, from a chat window to autonomous agents, and the stack underneath: models, access, protocols, frameworks, runtimes, governance and where it runs. Switch on Where MERLIN sits to see what MERLIN uses, what it leaves out by design, its gaps and where it could move. Every example links to its source.
The Templates tab has every message between the blocks, captured from a real run, and the catalogues: the list on the left (click a block to fold or unfold it), the message in the middle, and where it travels and the checks made against it on the right.
The inspector on the right closes with ✕, Esc or a click anywhere outside it.