Choose a path.
Classify intent, select a workflow, route a ticket, or choose among tool options. Keep the full distribution, including an escalation label when you need one.
Seb Labs for Vercel AI Gateway
Small model.
Clear decisions.
Route a request. Check the evidence. Score against a rubric. Seb-9B-v2 returns a typed judgment and its probabilities, ready for the next step in your application.
Direct API and MCP available today. Gateway listing requested.
A judgment for
the job at hand.
Supply the facts and a bounded question. Receive structured output without a generated explanation.
Check a proposition.
Estimate the probability of a clearly defined yes/no statement. Screen evidence, relevance or policy fit using the context you provide.
Apply your rubric.
Evaluate urgency, relevance or quality across ordered levels. Receive an expected ordinal score and the probability of each level.
Probabilities are model estimates. Apply your application’s uncertainty policy and existing controls before consequential actions. Use code for exact checks and retrieval for missing facts.
Build with Seb today.
Create a named inference token in the developer console and keep it in your environment as SEB_API_KEY. The direct endpoint runs the live model and shares your account balance with MCP.
Current Seb pricing is $0.021 per million input tokens, with output free. Each question meters its state, instructions and criteria. Failed or invalid inference is not billed.
For coding agents, the Claude Code and Codex plugin installs MCP and a decision skill together. The free seb_demo checks setup using illustrative data.
curl https://mcp.seblabs.dev/api/v1/systemone \
-H "Authorization: Bearer $SEB_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"state": "The parcel arrived damaged. Please replace it.",
"questions": {
"team": {
"type": "choice",
"instructions": "Choose the team to handle this request.",
"criteria": {
"returns": "Damaged products and replacements",
"billing": "Payments and invoices",
"escalate": "Unclear or outside these teams"
}
}
}
}'The Gateway path.
Can I call Seb through AI Gateway now?
Not yet. We have requested provider onboarding. Adding our MCP server, a BYOK credential or a private model alias does not publish Seb in the Gateway catalog. Use our direct API or MCP while this work is pending.
What integration are you pursuing?
A native decision-model route: your application → Vercel AI Gateway → Seb. Vercel already supports decision models; the provider protocol, qualification, capacity, billing and catalog publication must be agreed with its team.
Will the direct API keep working?
Yes. Direct API and MCP access are the current product. A future Gateway route would provide another way to reach Seb, with its final model ID, availability and pricing published once confirmed.
Where are the weights and evidence?
The Seb-9B-v2 model card contains the open weights and research materials. Hosted inference currently uses a text-only NVFP4 serving variant. Read our evaluation notes for benchmark scope, source differences and limitations.
Give your agent
a decision skill.
Install MCP, keep the key private, and make the next judgment explicit.