agynd

The agynd daemon bridges agent CLIs with the platform by connecting to Threads and Notifications services, preparing the agent runtime environment, and managing the agent process lifecycle.

Architecture: https://github.com/agynio/architecture/blob/main/architecture/agynd-cli.md

Local Development

Full setup: https://github.com/agynio/architecture/blob/main/architecture/operations/local-development.md

Run from sources

agynd is not deployed as a service. The Runner starts it as the main process of an agent container, so there is no Deployment to attach to and no devspace dev to run -- the way to exercise a change is to put the binary in an agent init image and point an agent at it:

CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -o .build/agynd ./cmd/agynd

Then build an init image that copies .build/agynd over the one baked in, load it into the local VM, and set it as the agent's init image:

agyn local load-image my-agent-init:dev

E2E validation

The GitHub E2E workflow runs this repository's local E2E tests with:

go test -v -count=1 -tags e2e ./test/e2e/

Those tests validate the local agent CLI bridge behavior against deterministic TestLLM endpoints through the Codex and AGN SDK flows. The workflow checks out agynio/agn-cli so the AGN coverage builds and runs the current AGN CLI during the test. They also build and execute this repository's cmd/agynd binary with /agyn/config.json installed by the test harness. That test uses a stub Gateway and fake AGN agent to verify daemon startup, platform initialization, and subscriber startup without requiring a full cluster.

This repository does not run the centralized agynio/e2e smoke suite. Broader platform and service smoke coverage remains owned by the centralized E2E repository and service-specific workflows.