Project-tracker agent Kaiku scopes itself to the calling user's permissions and logs every run; discloses no compliance certification or insurance

Kaiku, an AI agent built into a project-tracker and wiki product, runs with the permissions of the person who invoked it rather than a standing account of its own, according to the company's own site.

"An agent is a role of your workspace, not an account: no password, no token, no access of its own," the company states. "It runs as the person who called it, with the permissions they have at that moment, and is clamped to the issue's project — a question about one project never ships the other nine to a model." The company also says every agent action is kept: "Every agent run is recorded against the issue it worked on: tokens by kind, machine time, who asked." Kaiku describes itself as advisory rather than autonomous: "An agent proposes; a person approves," and "It never decides."

AI Dispatch's trust-gap rubric, scored against Kaiku's published material Oct. 13, credits the scope and logging controls described above: runtime governance scores 2 of 3 and audit trail scores 2 of 3, both short of a published, customer-reviewable retention or tamper-evidence guarantee. Compliance, outcome settlement and insurance-indemnity each score 0 of 3 — Kaiku publishes no compliance certification, no outcome-linked pricing (its plans are flat per-seat fees, starting at $19.99 a month for a single user), and no insurance or indemnity terms. A full-text search of Kaiku's homepage, a 13,285-character capture retrieved Oct. 6, found no occurrence of "insurance," "indemnif," "liabilit," "certif," "arbitrat," "warrant," "SOC 2" or "ISO 27001."