Documentation

The Polymer API serves the material channel of the genome — stacking energy, curvature, flexibility, groove geometry — computed at base-pair resolution across 50 layers. Every response includes api_version, data_version, an evidence class (M / K / D / S / H), source URI, license, and a content hash so agents and downstream pipelines can audit and cache.

Install the SDK and fetch your first region. The example uses the TP53 locus on hg38.

pip install polymer-genomics

from polymer_genomics import PolymerClient

client = PolymerClient(api_key="POLY_...")

region = client.regions.fetch("hg38", "chr17:7668421-7687490")
print(region.stacking_dg37.mean)            # -8.42 kcal/mol
print(region.cpg_sites.count)                # 47
print(region.evidence_classes)               # {'thermodynamics': 'D', 'cpg_sites': 'K'}

Without an SDK, hit the REST API directly. Use the curl below — any HTTP client works.

curl 'https://api.polymerbio.org/v1/genes/hg38/TP53' \
  -H 'X-API-Key: $POLYMER_API_KEY'

All requests must include the X-API-Key header. Request a key at polymerbio.org/developers. Free tier: 1,000 requests/day. Rate limits are returned in X-RateLimit-Remaining.

Reference endpoints return curated data with provenance. Compute endpoints run the physics linter or aggregate over a region. All responses include data_version and content-hash headers.

curl https://api.polymerbio.org/v1/regions/hg38/chr17:7668421-7687490 \
  -H 'X-API-Key: $POLYMER_API_KEY'

from polymer_genomics import PolymerClient

client = PolymerClient(api_key=POLYMER_API_KEY)
region = client.regions.fetch("hg38", "chr17:7668421-7687490")
print(region.stacking_dg37.mean)

curl https://api.polymerbio.org/v1/genes/hg38/TP53

gene = client.genes.fetch("hg38", "TP53")
print(gene.constraint.pli, gene.constraint.loeuf)

curl 'https://api.polymerbio.org/v1/aggregation/hg38/chr17:1-83257441?layers=cpg_sites'

agg = client.aggregation.fetch("hg38", "chr17:1-83257441", layers=["cpg_sites"], resolution=1_000_000)

curl https://api.polymerbio.org/v1/probes/hg38/cg13580121

probe = client.probes.fetch("hg38", "cg13580121")
print(probe.position, probe.context)

curl https://api.polymerbio.org/v1/layers

layers = client.layers.list()
for l in layers:
    print(l.id, l.version, l.evidence_class)

curl 'https://api.polymerbio.org/v1/search/genes?q=TP53'

hits = client.search.genes("TP53")
for h in hits:
    print(h.symbol, h.chromosome)

curl -X POST https://api.polymerbio.org/v1/evaluate \
  -H 'Content-Type: application/json' \
  -d '{"sequence":"ATGCGATCGATCG...","analysis":"full"}'

result = client.evaluate("ATGCGATCGATCG...", analysis="full")
print(result.summary.gc_content, result.flag_counts.warnings)

The Model Context Protocol server exposes 70 tools to LLM agents — 38 reference and 32 compute. Connect via stdio for local agents or HTTP/SSE for hosted deployments.

# Local stdio (Claude Desktop, etc.)
uv run polymer-mcp serve

# Hosted HTTP/SSE
polymer-mcp serve --transport sse --port 8050

All errors return a JSON body with code, message, and optional details. Status codes follow standard HTTP conventions.