ZCode
Zhipu's open-source workbench for running AI agents on a codebase
Emerged finds new products worth a look, daily — what’s new today
ClientCoded
Check an AI agent's answers against computed ground truth, not another LLM
Teams shipping data agents get a test set with SQL-computed answers, so they can catch wrong replies before customers do instead of trusting an LLM judge.
ClientCoded tests and monitors AI agents that answer questions over business data. It ships 40 synthetic environments modelled on tools like Salesforce, Jira and Stripe, each with about 200 adversarial queries whose correct answers are computed by SQL. It also generates personas that push agents off-script, and scores live conversations with Slack alerts. The first test is free; plans run $49 to $599 a month. Built by two bootstrapped founders, one from Veeva QA.
More like this
Xencode
A Rust terminal coding agent that defaults to models on your own machine
Canary
Your coding agent hands its changes to Canary, which attacks them in sandboxes
Thrum
A local message bus that lets your coding agents review each other's work
Permaura
A credential vault and approval gate between AI assistants and your apps
DashClaw
Approval gates and an audit ledger for coding agents left running unattended