Preparing components...
Skip to content

Reference benchmark replay and recovery simulation. No live database is killed. Current product results: 700 mean tokens over seven fixture tasks.

Fewer tokens.
Right order.

Recorded recovery demonstration, not a promise for this implementation.

110,238

tokens to hand an agent all 420 tools from 20 MCP servers, before it reads a word of the task.

Counted with the cl100k tokenizer over the real tool schemas.

Top-left wins.

Each dot is a way of picking tools. Further left means fewer tokens in context. Higher means more of the right tools, in an order that runs.

Reference replay only. Current implementation is measured on seven local fixture tasks; no live latency is being reported here.

1001k10k100k0 %25 %50 %75 %100 %Tokens in context, log scale. Fewer is better.Chain recall. Higher is better.Where you want to bepreload-allBM25 top-5BM25 top-10Switchboard

Reference comparison data from the original demo. These values are not current HydraDB measurements.

MethodChain recallOrderDisambiguationCalls savedRound tripsTokens
preload-all100.0%0.0%20.0%0.000.0110,238
BM25 top-531.0%0.0%0.0%0.001.01,430
BM25 top-1062.0%0.0%20.0%0.001.02,471
Switchboard100.0%100.0%100.0%1.001.0469

No current latency data is connected. The grid and kill flow below are an illustrative replay, not a live HydraDB benchmark.

Dark forest with light through the trees

Pull the plug. Watch the line.

Reference replay only. Current implementation is measured on seven local fixture tasks; no live latency is being reported here.

p50 …p95 …p99 …

Reference replay only. Current implementation is measured on seven local fixture tasks; no live latency is being reported here.

Recorded run, 140 plans: p50 269 ms, p95 295 ms, p99 307 ms.

Send a few plans

Kill the database

Local demo only.

DO NOT PEEL
peel to arm
Status ALIVE
Graph intact
  • Last recorded kill: plans worked again after 3.3 s, full recovery took 41 s, and the graph came back with the same counts.

What's in the graph.

Every tool comes from a real server's tools/list response.

420 tools across 20 servers

How it's measured

Search Switchboard

Search the docs, plan a task, or jump to a page