<!-- Generated by scripts/generate-llms.mjs from docs/connectors/fred.md. Do not edit. -->

Source: https://qbash.com/docs/connectors/fred

# Connect FRED

Tasks call FRED (Federal Reserve Economic Data, from the Federal Reserve Bank of St. Louis) through `qbash.integrations.fred` with an API key you store on the Connectors page. The connector searches FRED's economic time series by keyword or tag and reads their observations, with FRED's transforms such as percent change from a year ago. It also browses FRED's categories and releases and reads the release calendar. Every method is a read.

After you connect FRED, you can add its tools to a [qbash MCP server](https://qbash.com/docs-markdown/mcp-servers/overview.md#package) for ChatGPT, Claude and other MCP clients, or call it from a Python task. The [FRED MCP server](https://qbash.com/connectors/fred) page lists what the connector can do.

This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.

## Request an API key

1. Sign in at [fredaccount.stlouisfed.org](https://fredaccount.stlouisfed.org), or create a free account there.
2. Open **API Keys** and request a key. FRED asks for a short description of the application that will use it.
3. Copy the key, a 32-character string of lowercase letters and digits.

FRED asks for a separate key for each application, so request one for qbash instead of reusing a key from another tool.

## Add the credential in qbash

1. Open **Connectors**, then **FRED**, and click **Connect**.
2. Give the credential a name and paste the API key. qbash checks the key with FRED before saving it.
3. In the task editor, choose **Add connector** from the Package **+** menu and select the credential. See [Task resources](https://qbash.com/docs-markdown/tasks/task-editor.md#resources).

To edit the credential later, leave the key blank to keep the saved one.

## Call FRED from a task

This task reads CPI inflation since 2020, as the percent change from a year earlier, and logs the peak and the latest month.

```python
cpi = qbash.integrations.fred.get_observations(
    series_id="CPIAUCSL",
    units="pc1",
    observation_start="2020-01-01",
    sort_order="asc",
)

rows = [row for row in cpi["observations"] if row["value"] is not None]
peak = max(rows, key=lambda row: row["value"])
latest = rows[-1]

qbash.log(f'CPI inflation peaked at {peak["value"]:.1f}% in {peak["date"][:7]}.')
qbash.log(f'The latest reading is {latest["value"]:.1f}% for {latest["date"][:7]}.')
```

Use `search_series` to find a series ID, such as `UNRATE` for the unemployment rate or `FEDFUNDS` for the federal funds rate. The [SDK Reference](https://qbash.com/docs-markdown/tasks/sdk-reference.md) lists every method with its parameters and return values.

## Compare two series

Pass the same window and frequency to each series, so the dates line up.

```python
window = {"observation_start": "2015-01-01", "frequency": "m", "sort_order": "asc"}

unemployment = qbash.integrations.fred.get_observations(series_id="UNRATE", **window)
fed_funds = qbash.integrations.fred.get_observations(series_id="DFF", aggregation_method="eop", **window)

rates = {row["date"]: row["value"] for row in fed_funds["observations"]}
for row in unemployment["observations"][-12:]:
    qbash.log(f'{row["date"][:7]}: unemployment {row["value"]}%, fed funds {rates.get(row["date"])}%')
```

## Read the release calendar

`get_release_dates` starts from today by default, so it lists what FRED expects to publish next, soonest first. Pass a `release_id` from `list_releases` for one release, such as 10 for the Consumer Price Index.

```python
upcoming = qbash.integrations.fred.get_release_dates(release_id=10)
for entry in upcoming["release_dates"]:
    qbash.log(f'CPI release scheduled for {entry["date"]}')
```

## Add test data

FRED is public data, so a new key reads every series at once and needs no setup. `CPIAUCSL` (CPI), `UNRATE` (unemployment rate), `GDP` and `DGS10` (10-year Treasury yield) are useful series to test with.

## Details that affect results

- `get_observations` returns the newest 1,000 values first by default. Set `sort_order="asc"` for the oldest first, and narrow a long daily series with `observation_start` or a lower `frequency` instead of a large `limit`. The limit goes up to 100,000, and chat tools cap it at 1,000.
- Values come back as numbers. FRED writes a missing value, such as a market holiday, as `.`, and the connector returns it as `None`.
- `units` transforms the values: `pch` is the percent change from the previous period, `pc1` the percent change from a year ago and `pca` the compounded annual rate. `frequency` aggregates to a lower frequency with `aggregation_method` `avg`, `sum` or `eop` (end of period). FRED returns an error for a frequency higher than the series' own.
- With `frequency` set, FRED counts the observations before aggregating, so `count` can be far higher than the rows a call returns. Page with `next_offset` until it is `None`.
- `realtime_start` and `realtime_end` read the values as FRED published them at the time, from ALFRED, the St. Louis Fed's archive of past data vintages.
- FRED allows 120 requests per minute for each key. The connector retries a rate-limited call once.