Free recipes / Safety, Costs and Troubleshooting

Cut Your Hermes Agent Token Costs

A message that reads your real prompt-size and usage numbers and returns a ranked plan to cut Hermes token costs, without changing anything until you agree.

Your provider bill is higher than you expected, simple questions feel expensive, or you want a cost baseline before you add more skills and MCP servers. Saves about 40 minutes

The recipe

Role: You are a frugal, honest advisor on my own Hermes Agent setup. Context: Fixed prompt budget from hermes prompt-size: [PROMPT_SIZE_OUTPUT] My usage figures, from hermes insights or /usage: [INSIGHTS_OUTPUT] My main model is [MAIN_MODEL]. I mostly use Hermes for: [MAIN_USES]. Goal: Cut my token costs without making Hermes less useful for those uses. Steps: 1. From the prompt-size output, name the three biggest fixed costs (tool schemas, skills, SOUL.md, memory, context files) with their token counts. 2. Toolsets: list enabled toolsets I probably do not need for my uses, to switch off with hermes tools per platform. Keep memory and session_search. 3. MCP: for each connected server, suggest a tools.include list of what I actually use. 4. Skills: point out skills I may never use, and any pinned with skills.auto_load that should not be. 5. Routine work: where it fits, suggest cron.model for scheduled jobs, delegation.model for subagents, auxiliary.compression.model for summaries, and --no-agent for jobs that only run a script. 6. Habits: one model per session, because each /model switch resets the prompt cache; /new when the topic changes; /compress in long sessions. Rules: - Plan only. Do not change config or run anything. - Do not invent prices or savings. When an effect is a guess, say so. - Never suggest turning off approvals or memory to save tokens. Output format: a table with Change, Where (exact command or config key), Expected effect, Downside, ranked by effect. Then Do first: the top three, and how to measure the result with hermes prompt-size next week.

Fill in these brackets

`[PROMPT_SIZE_OUTPUT]`	paste the full output of hermes prompt-size
`[INSIGHTS_OUTPUT]`	paste the output of hermes insights, or /usage from a typical session
`[MAIN_MODEL]`	e.g. the model id shown by /model
`[MAIN_USES]`	e.g. morning brief, email drafts, research on my phone, some coding

Pro tip

Run hermes prompt-size and save the output before you change anything. Without a baseline you cannot tell whether a change helped, and the fixed cost is paid on every single call.

Variation

A week after the changes, send this again with fresh numbers and ask Hermes to compare the two runs line by line, so you keep only the changes that actually made a difference.

Written for Hermes Agent v0.21.x. Check commands against your version with hermes --version. This site is independent and not affiliated with Nous Research.