Liquid Inference update: reservations, MCP server and spend guardrails
Liquid Inference shipped monthly token reservations, workload testing to pick a provider, shared org budgets, a one-command MCP server and per-end-user spend limits.
Shipped this week on Liquid Inference: • Reservations: a monthly token pool from a provider at a fixed fee • Workloads: test a task on your own cases, then buy the provider that wins • Organizations: one balance and one budget for all your team workspaces • MCP server: add it to your agent with one command • Auto: keeps a session on the provider that holds its cache • Guardrails: new rate limits, and a spend budget for each of your end users • Speech to text: priced per second of audio • http://liquid.free: buys only free offers • Add :web to any model to search the web first Try it all on http://liquidinference.ai