The agentic rag, wrung out.
The rag for
The production-hardening of multi-agent and multi-profile setups
The dominant theme across high-signal posts is no longer “can agents do the work?” but “how do we run them safely, portably, and governably at scale?” - Hermes users are debating real gateway topologies and isolation boundaries. - Long-form guides are crystallizing the correct way to treat an agent’s brain as versioned files rather than a running process. - New tools (Campfire) are explicitly targeting the governance and audit gaps that pure throughput orchestrators leave open, while OpenClaw continues to receive security hardening (HSM secrets, enterprise control plane patterns). This is the shift from experimental agent harnesses to operational infrastructure. It is the story that will matter most to serious builders in the next 30–60 days.
Campfire addresses the gap left by throughput-focused orchestrators (Herdr, etc.). It runs Claude Code, Codex, Goose, OpenClaw, OpenHands and others side-by-side. Risky tool calls become votes (majority, any-deny, owner-decides) with a 30-second window. Every session is a replayable JSONL protocol log. Local vector memory stores decisions and patterns per repo/session so the next session starts warm.
Long-form practical guide on making a self-hosted agent (Claude Code / Hermes-style) portable across machines. Documents three failed approaches (always-on gateway, rsync copy, declarative machine rebuild) and the one that held: private git repo of the agent’s home folder (settings, hooks, skills, memory) with end-of-session commits. Includes a 5-step checklist, what cannot be carried (logins, packages, assumptions), and a read-only audit prompt to hand to your agent.
Hermes Agent has landed on iOS and Android in open beta. Aimed at small-business operators who need to approve follow-ups, fix drafts, or check numbers from a phone without opening the laptop. The agent continues working at the desk; sign-off happens anywhere.
1ClawAI shipped HSM-backed secret storage, automatic redaction, scoped identities, short-lived tokens, and transaction previews specifically for OpenClaw agents that handle wallets or API keys.
Detailed trade-off analysis of running one multiplexed gateway (gateway.multiplex_profiles true) versus a separate gateway per agent profile. Covers failure domains, cron duplication, blast radius of runaway loops, log isolation, and the real meaning of isolation (process count is not a security boundary). Recommendation: multiplex unless a profile has its own channel, needs pinned runtime, separate user/limits, or must survive others’ restarts.
Hermes Agent can now automatically clear ended chats that have been idle past 90 days at startup (off by default since v0.22.3). When disk space is low it asks once instead of deciding for you. Pinned and open chats are never touched. Configurable via retention_days. No plugin required.
Never miss a night
One feed for everything, or one per agent harness you run. Rebuilt every night, with full credit on every item.
