Coder: a workspace for agent-assisted development
Coder provides self-hosted cloud development environments and AI coding-agent workflows. The useful AIWave angle is a controlled workspace-side assistant call with the same OpenAI-compatible base URL used elsewhere in this docs set.
Why this project
It was on the weekly GitHub Trending page on 2026-09-25. The official README says Coder is for cloud development environments and AI coding agents, and describes bringing models through its agent workflow. We selected it because a workspace is a natural place to run a small AIWave helper without moving customer keys into the editor.
Sources: official repository and weekly Trending list, both checked 2026-09-25.
Run a small workspace helper
python -m pip install openai
export AIWAVE_API_KEY="your-aiwave-key"
python - <<'PY'
from openai import OpenAI
import os
client = OpenAI(base_url="https://aiwave.live/v1", api_key=os.environ["AIWAVE_API_KEY"])
r = client.chat.completions.create(
model="deepseek-v4-flash",
messages=[{"role":"user","content":"Summarize the latest test failure in two bullets."}],
max_tokens=120,
)
print(r.choices[0].message.content)
PYRun this inside a Coder workspace after checking the workspace’s network policy. For native Coder model configuration, follow the project’s current AI Gateway documentation; do not assume a custom endpoint field exists in every release.
Cost estimate
| Assumption | Estimate |
|---|---|
| 1,500 input + 300 output tokens | about $0.0015 per helper call on deepseek-v4-flash |
| Rate source | AIWave pricing, effective 2026-08-27; checked 2026-09-25 |
Guardrails
- Keep the key in the workspace secret store or environment, not a template.
- Send only the error excerpt needed for diagnosis.
- Use request recovery for API failures.