News

What shipped, and why.

Every post is about something that shipped in the Lobstack app or its API, or a claim of ours that turned out not to be true. The changelog is the shorter read.

Latest ·

Start from the job, set the rules

Lobstack 0.11.18: new agents start from what you want done, every agent gets its own approval rules, and every reply says which model answered and why.

Read the post
Diagram from the post: you pick a job, a ready agent is made, its rules are checked from the top, and each reply names the model that answered.

Recent

How to Build a Local AI Agent Without Cloud APIs

Choose hardware, a tool-calling model, and a local runtime, then wire an agent loop with validation, approval gates, and evaluation.

Read

Provider price + 15%

Included usage is now billed at the provider's price. Past it, models cost provider price + 15%. Free is $5 to start, then your own key or Copilot.

Read

Why we are Lobstack ASI

ASI is the direction AI is going. Lobstack builds the part that keeps a person in charge, and every model now comes from a US company.

Read

Lobstack is code-signed

From 0.11.10 the Windows installer is signed and the Mac app is notarized. The Mac warning is gone; Windows may ask until the certificate earns reputation.

Read

Archive

19 posts · since 8 September 2026
  1. Provider price + 5%We cut the Gateway rate from 25% over the provider's price to 5%. A plan now pays for Lobstack, and the models are passed through near cost.
  2. x402: agents that pay per requestA new x402 Payments Agent template and five skills: an agent finds a pay-per-call API, pays a few cents inside your rules, and shows the receipt.
  3. Links open in Lobstack’s own browserA link in an agent's message used to take you out of the app. It now opens in a window of the app's own; ⌘, Ctrl or Shift-click opens your browser.
  4. A Lobstack agent made its first on-chain paymentOne sentence in a chat, 0.5 USDC on Monad testnet: planned, approved, sent and checked against the chain in 44 seconds.
  5. Plan, Act, Verify: how an agent pays without being trusted with your moneyAgent payments in Labs give an agent its own wallet on Monad testnet. The agent can spend it, but it cannot change the rules or skip a step.
  6. The canvas has tabsFiles, an agent's browser, automations and runs open side by side next to the chat. Designs show at their real size, and you can point at one element.
  7. An agent's browser you can watch, drive and replayEvery click lands with a ripple and a named step. Take Over hands you the whole page, and an agent stuck at a sign-in now stops and asks for you.
  8. Workflows, and a dry run that says what it would have doneA routine can now have steps, each with its own agent, tools and approval. A dry run does the real look-ups and lists every write it would have made.
  9. “I don't have access to the add_rows tool”An agent kept the narrow tool list of a job that failed while planning, and handed the work back. The fix, and an audit of the tools behind every feature.
  10. Approve from your phone, through a relay that cannot read the cardWhen an agent stops to ask, a phone you paired can answer. The card is sealed on your computer, and lobstack.ai carries it without being able to read it.
  11. The desktop app is now called LobstackThe desktop app takes the company's name in 0.8.0. Your agents, threads, keys and settings stay exactly where they were.
  12. Sign in with Lobstack, and $5 to startFrom 0.7.0 a new install opens on one button. Your agents then run on the Lobstack API, and your first sign-in adds $5 of credit.
  13. A terminal UI for the CLI, still with zero dependenciesThe CLI opens a full-screen UI showing what each call cost as it happens, drawn by hand because a key-holding process should not pull a package tree.
  14. An Agent Skill is not a connectorThe Skills page listed 153 credentialed third-party integrations and called them skills. Not one of them was an Agent Skill.
  15. Plans include dollars of model spend, not messagesA message is not a unit of cost. Every previous plan was underwater at its own advertised limit on any flagship model.
  16. Comparison pages, with a build check that fails when a claim goes staleFour gateways, one typed fact table, a first-party source on every claim, and a sixty-day shelf life the build enforces.
  17. The price rides on the frame that carries the tokensA streamed response cannot carry cost in a header, so the last chunk before [DONE] now carries it under x_lobstack, beside the usual usage object.
  18. The Gateway returned 400 to every OpenAI request for three monthsOne sampling parameter the caller never asked for, roughly 1,395 consecutive failures, and three months of blaming the provider credentials.
  19. Every Gateway call comes back with a receiptCost used to be recomputed at read time from today's price list. It is now priced when the request is served and frozen on the ledger row.