INTEGRATION / AGENT RUNTIMES

Self-improving agents, audited every step.

Hermes agents learn: they create skills from experience, persist memory across sessions, and run scheduled jobs unattended. That power needs governance. Obsivara audits what your agent learned and when, flags sensitive data written to memory, catches skill regressions, and alerts you when the 9 AM report job silently stops running.

Hermes AgentEXAMPLE
AGENT RUNS TRACED / 24H2,417
TOOL CALLS MONITORED / 24H9,338
HIGH-RISK ACTIONS FLAGGED / 24H23
ILLUSTRATIVE · EXAMPLE DATA
WHAT YOU GET

Hermes Agent, fully observable.

Memory governance

Every memory read, write, and delete tracked. Sensitive writes are flagged high-risk so PII never quietly lands in long-term memory.

Skill audit trail

Skills tracked like assets: created, updated, executed — with version history, success rates, and regression detection when a skill update degrades performance.

Scheduled job monitoring

Missed, late, and failed cron jobs surfaced with business impact — know the executive report didn't go out before the executives do.

Provider-routing cost intelligence

Hermes routes across providers; Obsivara attributes cost per skill, per job, and per model — with savings recommendations.

HOW TO CONNECT

Three steps. No code changes.

1STEP 01

Create an Obsivara API key

Generate an ingest key in Settings → API Keys.

2STEP 02

Connect your Hermes instance

On the Integrations page choose Hermes Agent, paste your API server URL and key, and select the agents to monitor.

3STEP 03

See the learning loop

Runs, skills, memory events, and scheduled jobs stream into the Agent Runtimes page within one sync cycle.

WHAT WE MONITOR

Signals tracked out of the box.

  • Agent runs across CLI, gateway, scheduler, and API triggers
  • Skill creation, updates, and per-version success rates
  • Memory reads/writes with sensitivity flagging
  • Scheduled job punctuality and failures
  • Provider and model routing decisions
  • Token usage and cost per skill and job
INTEGRATION FAQ

Common questions.

Does Obsivara see my memory contents?

No — memory events carry keys, operations, and sensitivity flags, not stored values.

Can I get alerted when the agent creates a new skill?

Yes. Skill-created events appear in the activity feed and can raise recommendations for review before broad use.

What happens when a skill regresses?

Obsivara compares success rates across skill versions and raises a recommendation to review or roll back the change.

Connect Hermes Agent in minutes.

No credit card · 5-minute setup