How to Use GLM-5.3 in OpenClaw
GLM-5.3 is built for complex coding, debugging, and tasks that require many steps. Compared with GLM-5.2, Z.ai reports stronger coding and extended-task performance.
Prompts to try with GLM-5.3 in OpenClaw
Copy an example into OpenClaw, or ask in your own words.
Each morning, use GLM-5.3 to analyze the supplied error summaries. Group recurring failures and cite evidence for new patterns.
Each day, use GLM-5.3 to compare supplied benchmark measurements with the baseline. Flag regressions and propose follow-up experiments.
Each week, use GLM-5.3 to review supplied code changes for security flaws. Include supporting code excerpts and proposed fixes.
After each test run, use GLM-5.3 to analyze the supplied failure text. Separate likely regressions from failures already recorded.
Each week, use GLM-5.3 to turn supplied engineering updates into JSON with completed work, blockers, and next actions.
How to set up GLM-5.3 in OpenClaw
Pick the way that suits you. Each takes about two minutes.
Use the Spiral skill
- Send OpenClaw this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
Use Spiral's MCP server
- Add the server
On the machine that runs OpenClaw's Gateway:
openclaw mcp add spiral \ --url https://api.spiral.ai/mcp \ --transport streamable-http \ --auth oauth
- Sign in
In Settings › MCP, select Sign in on the Spiral server, sign in to Spiral and click Authorize. No API key.
openclaw mcp login spiral
- Reload and check
A running Gateway may keep its old tool list: restart it, or Save & Publish in the Control UI.
openclaw mcp reload openclaw mcp doctor spiral --probe
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run GLM-5.3
One request; the result comes back by polling or a webhook.
curl https://api.spiral.ai/v1/runs \ -H "Authorization: Bearer $SPIRAL_API_KEY" \ -H "Content-Type: application/json" \ -H "Idempotency-Key: $(uuidgen)" \ -d '{ "provider": "openrouter", "model": "z-ai/glm-5.3", "input": { "messages": [ { "role": "system" } ] } }'
Then paste one of the prompts above: OpenClaw finds GLM-5.3 on Spiral, shows the price and runs it.
GLM-5.3 providers and prices
Spiral runs GLM-5.3 through OpenRouter. Open a card for today's price, with every fee included.
Up to $0.007459 per run at the default inputs. $0.09087 per 1M input tokens, $1.82 per 1M output tokens.
Settings you can ask for
Ask OpenClaw for any of these in your prompt.
| Setting | Allowed values | Default | Description |
|---|---|---|---|
| Seed | a whole number | ||
| Min p | 0 to 1 | ||
| Top k | a whole number from 0 | ||
| Top p | 0 to 1 | ||
| Logprobs | on or off | ||
| Max tokens | 1 to 943718 | The most output tokens to generate. | |
| Temperature | 0 to 2 | ||
| Top logprobs | 0 to 20 | ||
| Presence penalty | -2 to 2 | ||
| Reasoning effort | low, high, or max | max | How much the model reasons. |
| Frequency penalty | -2 to 2 | ||
| Repetition penalty | 0 to 2 | ||
| Parallel tool calls | on or off |
Every API field of Z.ai: GLM 5.3 is on its tool page: open its card above.
What can you do with GLM-5.3 in OpenClaw?
GLM-5.3 can analyze recurring technical evidence and produce a consistent report for review. Define what counts as a new issue and require evidence for each finding.
Repair a difficult bug
Ask GLM-5.3 to trace a failure, propose a minimal patch, and define a regression check.
Plan a code migration
Use GLM-5.3 to identify dependencies, order changes, and define compatibility checks.
Implement a feature
Give GLM-5.3 requirements and acceptance criteria, then request an implementation and verification report.
Investigate performance problems
Have GLM-5.3 examine code and measurements, propose a bottleneck hypothesis, and design a comparison.
Review code for vulnerabilities
Ask GLM-5.3 to inspect your code for security flaws and suggest fixes with supporting evidence.
Evaluate a technical design
Use GLM-5.3 to compare architectures against stated constraints and identify unresolved decisions.
Extract structured records
Ask GLM-5.3 to turn supplied text into JSON with explicit rules for missing information.
More on GLM-5.3 and OpenClaw
Overview
GLM-5.3 in other agents
More OpenClaw guides
Questions about GLM-5.3 in OpenClaw
Can OpenClaw use GLM-5.3?
Yes. Add Spiral to OpenClaw as an MCP server, and OpenClaw can find and run GLM-5.3 like any built-in tool.
How much does GLM-5.3 cost in OpenClaw?
Up to $0.019174 per run at the default inputs. $0.1092 per 1M input tokens, $4.68 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your OpenClaw plan is separate.
Do I need an OpenRouter account to use GLM-5.3 in OpenClaw?
No. Spiral runs GLM-5.3 on its own OpenRouter account. You sign in to Spiral once and never paste a key.
How long does it take to set up GLM-5.3 in OpenClaw?
About two minutes: add the server, then sign in, then reload and check.
Does OpenClaw wait while GLM-5.3 runs?
GLM-5.3 runs in the background. OpenClaw starts the run, checks on it and gives you the text, json when it's ready.
What is GLM-5.3 good at?
Complex coding and tasks that require sustained work across several steps are its main focus. Try it for debugging, refactoring, and implementing changes with explicit acceptance criteria.
Can GLM-5.3 understand screenshots or video?
GLM-5.3 accepts text only. Provide a written description, extracted text, or logs when the original evidence is visual.
How much text can GLM-5.3 handle?
Z.ai documents a 1 million-token context window and a maximum output length of 128K tokens. Keep requested answers focused even when supplying extensive context.
Can I turn off GLM-5.3's reasoning?
No. Thinking stays enabled; use low reasoning effort for simpler work and start with maximum reasoning for difficult problems.
Where does GLM-5.3 fall short?
Its coding results are not uniformly best: Z.ai reports that Claude Fable 5 scores higher on its Code Bench. Treat generated patches as candidates to test against your own requirements.
Start using GLM-5.3 in OpenClaw
One Spiral account for every tool your agent needs. No subscription; you pay per run.
Paste this into OpenClaw:
set up https://spiral.ai/SKILL.md


