Every session, your agent starts from zero.
Ours doesn't.
A guild gives each of your projects a named agent with memory of its own: decisions, gotchas, and open work carried across every session. It opens each conversation knowing who it is, what you did last time, and what's still open.
Setanta — guild memory loaded: 3 open tasks, last worked yesterday
I'm Setanta, this project's agent, carrying the memory of every session we've had here.
Where we left off:
· Programme visibility shipped and verified on prod
· SEO canonical fix deployed — commit 6b1ac60
Still open:
· Restore-test the backup cron
· Mobile audit follow-ups for the coach section
A guild is you, plus your agents
One named agent per project you choose to give memory. Some projects should remember; some shouldn't. You decide, and the roster is yours to run.
Memory the agent owns
Not a transcript dump. These are graded memories the agent writes with hindsight: decisions above notes, gotchas above chatter. Wrong memories get superseded, never edited; finished tasks get closed. The history is append-only and honest.
Walled by default, shared by choice
Each agent's memory is owner-walled, so one project's agent can never see another's. Facts you want every agent to know go to the guild's shared circle with one flag. The walls are enforced by the platform, not by prompt engineering.
Works where you work
Claude Code gets automatic recall at session start. Everything else, from Codex to any
shell-capable tool, uses the same yohanun CLI.
Same agent, same memory, any tool.
The rhythm is three commands
yohanun search "auth flow decision"
Recall before reinventing. The agent checks its memory the way you'd check with a colleague.
yohanun store "chose bcrypt, 12 rounds" --type decision
Store as work lands. The type matters: decisions outrank passing notes in recall.
yohanun consolidate
Before a session ends, the agent distils what mattered. It handled the whole session, so it's the one qualified to say what's worth keeping.
Five minutes to a founding agent
$ pip install https://app.yohanun.com/static/kit/yohanun.tar.gz
$ yohanun init
# your guild + founding agent, created
$ yohanun agent add myapp --matter project:myapp \
--project-dir ~/dev/myapp --wire
# next Claude Code session opens with memory
Full quickstart lives in your dashboard once you're in the beta.
Built on governed memory, not a vector dump
Guilds runs on the same Yohanun platform built for regulated work: tenant isolation as the hard wall, per-agent ownership enforced inside every query, memory lifecycle that supersedes rather than edits, and recall that weights what you're working on now. That includes picking an old project back up as if you'd never left.
How the architecture works →The beta is small on purpose
Guilds is free while in beta, and invite-only. We work closely with early guilds and the product improves from real use. It will eventually be paid, and beta testers will hear that from us long before it happens.
Ask for an invite