Upload files directly from your computer to Agent Context. Agent Context automatically detects the schema and makes your data queryable with SQL. No database or cloud storage required.
- Go to Agent Context in the dashboard
- Click Add Source → Upload File
- Drag and drop your file, or click to browse
- The schema is auto-detected — review and confirm
- Your data is immediately queryable
| Parameter | Required | Description |
|---|
| File | Yes | The file to upload (.csv, .tsv, .parquet, .json, .jsonl, .xlsx, .xls, .duckdb, .sqlite) |
| File Format | No | Auto-detected from extension. Override: csv, tsv, parquet, json, jsonl |
| Format | Extension | Notes |
|---|
| CSV | .csv | Auto-detects delimiter, headers, and column types |
| TSV | .tsv | Tab-separated values |
| Parquet | .parquet | Columnar format. Preserves types exactly. Best for large datasets. |
| JSON | .json | JSON array format. |
| JSONL | .jsonl | Newline-delimited JSON. One record per line. |
| Excel | .xlsx, .xls | Each sheet becomes a separate table |
| SQLite | .sqlite | All tables in the database are discovered |
| DuckDB | .duckdb | All tables in the database are discovered |
- Quick exploration — drop a CSV export and start querying immediately
- One-off analysis — upload a spreadsheet your team shared for a specific task
- Prototyping — test queries on local data before connecting a production database
- Small datasets — reference data, lookup tables, configuration data
For regularly updated data, consider connecting a database or cloud storage source instead.
| Issue | Fix |
|---|
| Schema not detected | Ensure the file has consistent column structure. For CSV, check that the delimiter is consistent. |
| Wrong column types | Try uploading in Parquet format for exact type preservation. |
| File too large | For files over 50 MB, consider uploading to cloud storage (S3, GCS, Azure) and connecting that way. |
