OpenStreetMap Planet Files

Use in your agent

Load an OSM minute diff into pandas and Parquet. Checked against the provider's documentation on Nov 1, 2026.

Access

• No account or key. Replication state and compressed minute diffs are public downloads.
• The minute-diff example is a small incremental slice; initial full-database replication still requires a planet snapshot.

Request

curl -O 'https://planet.openstreetmap.org/replication/minute/007/354/232.osc.gz'

Format

Gzip-compressed OSM change XML with create, modify and delete groups containing node, way and relation elements. Elements include IDs, versions, timestamps, changesets, optional coordinates and nested tags or members.

Load into pandas and Parquet

Needs Python 3 with requests, pandas and pyarrow.

"""Latest OpenStreetMap minute replication diff -> pandas -> Parquet."""
import gzip
import xml.etree.ElementTree as ET
import requests
import pandas as pd

BASE = "https://planet.openstreetmap.org/replication/minute"
HEADERS = {"User-Agent": "SignalLabCheck/1.0"}
state = requests.get(f"{BASE}/state.txt", headers=HEADERS, timeout=30)
state.raise_for_status()
sequence = int(next(line.split("=", 1)[1] for line in state.text.splitlines()
                    if line.startswith("sequenceNumber=")))
path = f"{sequence:09d}"
diff_url = f"{BASE}/{path[:3]}/{path[3:6]}/{path[6:]}.osc.gz"
r = requests.get(diff_url, headers=HEADERS, timeout=60)
r.raise_for_status()
root = ET.fromstring(gzip.decompress(r.content))
rows = []
for group in root:
    action = group.tag if group.tag in {"create", "modify", "delete"} else "unknown"
    for obj in group:
        rows.append({
            "action": action, "type": obj.tag, "id": int(obj.attrib["id"]),
            "version": int(obj.attrib.get("version", 0)),
            "timestamp": obj.attrib.get("timestamp"), "lat": obj.attrib.get("lat"),
            "lon": obj.attrib.get("lon"), "changeset": obj.attrib.get("changeset"),
            "tag_count": sum(child.tag == "tag" for child in obj)
        })
df = pd.DataFrame(rows)
df["timestamp"] = pd.to_datetime(df["timestamp"], utc=True)
df[["lat", "lon"]] = df[["lat", "lon"]].apply(pd.to_numeric, errors="coerce")
df.to_parquet("osm_minute_diff.parquet", index=False)
print(len(df), "changed objects", "sequence", sequence,
      df["timestamp"].min().isoformat(), "to", df["timestamp"].max().isoformat())

Run by us on 2026-11-01 (python 3.13.5 | pandas 3.1.0 | pyarrow 26.0.0 | requests 2.34.2). Result: 3502 changed objects sequence 7354232 2026-11-01T17:09:11+00:00 to 2026-11-01T17:10:10+00:00

Limits

• Download rate limit: Not stated.
• The checked full snapshot was 89 GB as PBF and 167 GB as compressed XML.
• Weekly snapshots, daily diffs and hourly or minutely diffs have different lag and storage costs.

Quirks

• Apply replication sequences in order and persist the last committed sequence; skipping or replaying a diff can make a local database inconsistent.
• Ways reference node IDs and relations reference nodes, ways or other relations, so a flat object table is only a staging representation.
• Delete records and some non-node objects have no latitude or longitude; preserve action, type, ID and version.
• The full-history files include older versions and deleted objects; the latest planet snapshot represents current data.

Terms and attribution

• Current OpenStreetMap planet data is licensed under ODbL 1.0.
• Public use requires OpenStreetMap attribution and notice of the Open Database License.
• ODbL share-alike conditions can apply to altered or derivative databases; older pre-12 September 2012 files use CC BY-SA 2.0.

Sources

• https://wiki.openstreetmap.org/wiki/Planet.osm: “Planet files are very large (tens of GBs) so please check the size before starting to download.”
• https://wiki.openstreetmap.org/wiki/Planet.osm: “A daily diff is generally about 40-80 MB, but can easily grow to 300-400 MB (both compressed) in exceptional cases.”
• https://www.openstreetmap.org/copyright: “You are free to copy, distribute, transmit and adapt our data,”
• https://osmfoundation.org/wiki/Licence/Attribution_Guidelines: “The historical forms of attribution “© OpenStreetMap contributors” or “© OpenStreetMap” are acceptable.”