Skip to main content
Kallima is an antibody design API. You send it VH/VL sequences; it returns humanized variants, predicted structures, stability analyses, immunogenicity scores, and antibody-antigen complex predictions. Each pipeline runs on GPU in the cloud — you submit a job and poll (or subscribe via webhook) for results. The API is designed for bioinformaticians and engineers at small-to-mid-size biotechs who want to embed antibody optimization into existing Python pipelines, Jupyter notebooks, and LIMS integrations without standing up their own compute infrastructure.

​
What you can do

EndpointWhat it runsTypical turnaround
GET /v1/meCredit balance + org identityinstant
POST /v1/humanizationsCDR grafting + back-mutation strategies~15–30 s
POST /v1/structure-predictionsImmuneBuilder ABodyBuilder2 / NanoBodyBuilder2~1–3 min
POST /v1/stability-analysesΔΔG, solubility, thermostability~1–3 min
POST /v1/immunogenicity-analysesMHC-I/II + B-cell epitope risk~1–3 min
POST /v1/complex-predictionsBoltz-2 antibody-antigen docking~20–40 min
POST /v1/codon-exportsCodon-optimized CDS (FASTA / CSV / GenBank)~30–60 s

​
How the data model works

Everything lives under a project. A project holds source antibodies (your parental sequences) and therapeutic candidates (the design series you’re optimizing). Each candidate has one or more variants — a baseline variant is created automatically when you register a candidate, and you create new variants as you iterate. Pipelines run against a variant. 
project
  ├── source_antibodies    (parental sequences)
  └── therapeutic_candidates
        └── variants
              └── jobs     (humanization, structure, stability, immunogenicity, complex)
                    └── results

​
Base URL

https://api.kallima.bio
All endpoints require HTTPS.

​
Client libraries

Python SDK

pip install kallima — typed wrappers, auto-pagination, built-in polling.

REST API

Any HTTP client works — curl, httpx, requests, Postman.