Hermes SDK — Examples & artifacts

code agent.code("compute the first 20 Fibonacci numbers, print them and save fib.txt") 13.2s · 4 steps

Steps

✓edit write fib.py

✕execute python fib.py (retry)

✓execute python3 fib.py → stdout + fib.txt

Result

Printed the sequence and wrote it one-per-line to fib.txt.

0, 1, 1, 2, 3, 5, 8, 13, 21, 34,
55, 89, 144, 233, 377, 610, 987,
1597, 2584, 4181

Artifacts

📄 fib.txt · text/plain · 64 Bdownload

The code it wrote & ran — fib.py

fib = [0, 1]
for i in range(2, 20):
    fib.append(fib[-1] + fib[-2])
for n in fib:
    print(n)
with open("fib.txt", "w") as f:
    for n in fib:
        f.write(f"{n}\n")

🐍 fib.py · 175 Bdownload

charts agent.visualize("Quarterly revenue (k USD): Q1 120, Q2 150, Q3 135, Q4 190") 12.7s · 3 steps

Steps

✓execute run matplotlib

✓execute save quarterly_revenue.png

Result

A clean bar chart of quarterly revenue — title, axis labels, value labels, minimal styling — returned as a downloadable PNG.

Artifacts

quarterly_revenue.png · image/png · 38 KBdownload

image agent.image("640×360 PNG: warm gradient, centered text HERMES SDK") 19.1s · 5 steps

Steps

✓edit write PIL script

✕execute run (retry)

✓execute render → hermes_banner.png

Result

A 640×360 PNG: warm orange-to-dark gradient with HERMES SDK centered in white. Drawn with PIL.

Artifacts

hermes_banner.png · image/png · 2.7 KBdownload

🐍 hermes_banner.py · 1.2 KBdownload

browser agent.browse("open example.com and take a screenshot") 8.5s · 2 steps

Steps

✓browser open https://example.com

✓browser screenshot

Result

Drove a real Chromium to example.com and captured the page — the standard “Example Domain” page — as a PNG.

Artifacts

browser_screenshot.png · image/png · 15 KBdownload

speak agent.speak("This audio was generated by the Hermes SDK.") 11.9s · 4 steps

Steps

✓execute detect edge-tts

✓execute synthesize speech

✓execute encode → generated_audio.mp3

Result

A valid MP3 of the spoken text — generated via edge-tts. Press play. 👇

Artifacts

generated_audio.mp3 · audio/mpeg · 20 KBdownload

web agent.research("What is the Agent Client Protocol (ACP)?") 70.1s · 25 steps

Steps

✓web search the web

✓fetch read several pages

✓execute synthesize a grounded answer (25 steps total)

Result — grounded in live pages

The Agent Client Protocol (ACP) is a standardized protocol that enables communication between code editors/IDEs (Zed, JetBrains, Emacs) and AI coding agents (Claude Code, Gemini CLI, GitHub Copilot CLI, OpenAI Codex). Created by Zed Industries and JetBrains, modeled after the Language Server Protocol (LSP) — so any agent can work with any compatible editor, and vice-versa.

No file artifact — web returns a grounded answer in res.text.

Captured live from the Python API — from hermes_sdk import Hermes; agent = Hermes() — then each res.artifacts[i].bytes() written to the files linked above.