soapboxUpdated just nowSoapbox for AI agents
Everything on this site - the Index, the issue trends, the channel stances - is computed from a dataset of scored, quoted, source-linked issue mentions across high-reach political podcasts and YouTube shows. The charts answer the questions we thought to ask. Your questions are different, and there are more of them than any dashboard can hold.
So we expose the dataset directly to AI agents over MCP (Model Context Protocol). Connect Claude, Cursor, or any MCP-capable agent with one config block, and ask in plain English. The agent composes the queries; you get answers with verbatim quotes and source links.
Subscribe and we'll email you a link to set your password and connect your agent. Cancel anytime.
Who this is for
Campaign teams deciding what to answer and where. Media buyers placing spots against stance, not just demographics. Political consultants and pollsters who want a leading indicator on issue salience. Comms shops building clips memos at 6am. Journalists and researchers mapping who said what, when, with receipts. If your job involves knowing what political media is saying before everyone else does, this is the raw feed.
Ask it anything
These are real questions the dataset can answer today - each one resolves to a couple of tool calls your agent makes on your behalf:
“Build me a list of right-leaning podcasts under 1M reach whose tariff coverage softened in the last 30 days. Rank by audience and pull the quotes that show the shift.”
“What moved this week that we should be ready to answer on Sunday? Give me the three biggest swings with the sharpest quote from each side, with links.”
“Every quote about the shutdown fight with sentiment −3 or colder this week, formatted as a morning clips memo with source links.”
“Compare how independent and legacy media framed crime over the last 90 days. Where do the two curves diverge, and who's driving the divergence?”
“Which issues surged in mention volume in the last two weeks before showing up anywhere else? Plot the volume curves for the top five and tell me who lit each fuse.”
“Find middle-rated shows that discuss healthcare with right-aligned framing - those audiences may be movable. Quote the framing so I can hear the register.”
“Which M-lean channels gave the economy the most airtime this month? I want reach-ranked placement targets with each channel's stance profile.”
“What is the right talking about this week that the left isn't touching at all - and vice versa? Quotes and sources for both lists.”
What's underneath: nine tools
Your agent discovers these automatically on connect. Every tool is read-only.
| search_mentions | The workhorse. Filtered search over scored issue mentions - verbatim quotes with sentiment (−5 left to +5 right), intensity, channel, episode, and a source link. Ten filters, paginated. |
| get_issue_trend | Weekly time series for one issue: mention volume, average sentiment, average intensity. Trajectory questions. |
| get_index | The Soapbox Index (−10…+10) over any trailing window, with delta, daily sparkline, and top issues by volume. |
| get_movers | Issues with the biggest period-over-period lean or volume swings, with thin-sample noise filtered out. |
| get_issue_detail | One issue, drilled in: which channels drive it this week and from which side. |
| get_channel_detail | One show, drilled in: its issue mix and per-issue stance profile. |
| list_issues | The full issue taxonomy with definitions and the canonical left/right positions used in scoring. |
| list_channels | The tracked panel: every show with lean, cohort (independent vs legacy), platform, and audience reach. |
| get_methodology | How every number is computed, plus live panel statistics - so your agent can cite its sources. |
The data, and its boundaries
Every mention is a verbatim quote extracted at classification time, scored for sentiment (−5 strongly left-aligned to +5 strongly right-aligned) and intensity (1–5), and linked to its source episode. Aggregates use the same reach- and intensity-weighted math as the public site - the full derivation is on the methodology page, and your agent can pull it live via get_methodology.
One hard boundary: full transcripts are never exposed - through this API or anywhere else. You get mention-level excerpts and a link to the source episode. This is both a licensing obligation to our transcript providers and house policy.
Getting connected
Setup takes about two minutes and needs no coding: subscribe, set a password, and add Soapbox as a custom connector in your AI app (Claude, ChatGPT, Cursor, and more). It authenticates with a quick browser sign-in - there's no key to paste. The setup guide walks you through every click, with copy-paste configs for developer tools.