For agents

Your AI agent can read everything CivicFM's shows have reported: what each local government took up, how it voted, and what was said, item by item, each with a link to its moment in the recording. Connect it through the MCP server, or have it call the API.

An agent that finds the site by itself can start at /llms.txt, which says the same in Markdown and lists the shows.

MCP

Any AI agent that takes an MCP server's address can read every item here, and answer a question about your local government with links to the moments in the recordings. The server is https://civic.fm/mcp, read-only, and needs no sign-in.

• In Claude: open Customize → Connectors, choose Add → Add custom connector, name it CivicFM, paste the address, and Add.
• In Claude Code:

  claude mcp add --transport http civicfm https://civic.fm/mcp

Then ask it something like “What has Pasadena decided about surveillance technology lately?” It has 8 tools:

The API

What local governments decided at their public meetings, item by item, as JSON: what each body took up, how it voted, and what was said, each with a link to its moment in the recording. Items are filed by topic, across every show.

Everything it gives is on the shows' episode pages, produced with AI from the public recordings and checked against them. It's read-only and open: no key, and any site may call it. Version 1 only grows: fields are added, never changed or removed.

Right now it holds 723 items from 75 meetings, across 8 shows, from September 24, 2026 to October 30, 2026. It grows each time a show publishes its weekly episode.

Every call and field is listed below, and in the reference, where you can try each. Its description, as OpenAPI, is for a tool that writes a client or imports an API.

Try it

curl 'https://civic.fm/api/v1/items?topic=surveillance&q=drone&limit=1'

Open it in your browser. Right now it answers:

{
  "items": [
    {
      "id": "pasadena:13:2",
      "show": {
        "slug": "pasadena",
        "title": "Pasadena",
        "site": "https://pasadena.civic.fm"
      },
      "meeting": {
        "id": 13,
        "body": "City Council",
        "date": "2026-10-27",
        "title": "City Council Meeting: October 27, 2026 - Oct 27, 2026",
        "recording": "https://pasadena.granicus.com/MediaPlayer.php?view_id=25&clip_id=8902",
        "documents": [
          {
            "url": "https://pasadena.granicus.com/AgendaViewer.php?view_id=25&clip_id=8902",
            "kind": "agenda",
            "title": "Agenda"
          },
          {
            "url": "https://ww2.cityofpasadena.net/2026%20Agendas/Oct_27_26/Agenda.asp",
            "kind": "agenda",
            "title": "Council Agenda"
          }
        ]
      },
      "number": 2,
      "agenda_ref": "5",
      "title": "Police drone-operator training grant ($39,196), approved on consent",
      "what_and_why": "Consent item 5 authorizes the City Manager to execute documents for an Edward Byrne Memorial Justice Assistance Grant and amends the Police Department's FY2027 operating budget to recognize and appropriate $39,196 (agenda) in the General Fund. Staff said the money would train more police drone (UAS) operators to support SWAT; the item drew questions because the council has a pending discussion about expanding police drone use to 'drone as first responder' (DFR). It passed as part of the consent calendar motion covering all items except item 3. The clerk reported two letters asking staff to clarify the purpose and use of the grant funds.",
      "topics": [
        "surveillance",
        "policing",
        "contracts"
      ],
      "at": "00:07:50",
      "link": "https://pasadena.granicus.com/player/clip/8902?view_id=25&entrytime=470",
      "outcomes": [
        {
          "text": "A motion to approve all consent calendar items except item 3 was made and seconded. Result: passed.",
          "unclear": false,
          "at": "00:12:24",
          "link": "https://pasadena.granicus.com/player/clip/8902?view_id=25&entrytime=744",
          "result": "passed",
          "tally": null,
          "votes": null,
          "no": [],
          "abstain": [],
          "recused": []
        }
      ],
      "points": [
        {
          "text": "A councilmember said they had contacted the police chief beforehand because they shared public concerns about what the grant money would pay for",
          "at": "00:08:44",
          "link": "https://pasadena.granicus.com/player/clip/8902?view_id=25&entrytime=524",
          "speaker": {
            "name": null,
            "role": "a councilmember"
          }
        },
        {
          "text": "A councilmember said they are comfortable with SWAT's use of drones but not interested in expanding to drones as first responder",
          "at": "00:08:44",
          "link": "https://pasadena.granicus.com/player/clip/8902?view_id=25&entrytime=524",
          "speaker": {
            "name": null,
            "role": "a councilmember"
          }
        }
      ],
      "highlights": [
        {
          "title": "If the council rejects first-responder drones, where does this money go?",
          "context": "During the consent calendar, a member questions a police department representative about a grant to train more police drone operators, while the council still has a separate discussion pending on using drones as first responders.",
          "speakers": [
            {
              "name": null,
              "role": "a councilmember"
            },
            {
              "name": null,
              "role": "city staff"
            }
          ],
          "audio": "https://civicfm-demo-media.t3.tigrisfiles.io/pasadena/highlights/2026-10-30-95cd2ba5.mp3",
          "length_s": 61.1,
          "start": "00:10:56",
          "end": "00:11:57",
          "link": "https://pasadena.granicus.com/player/clip/8902?view_id=25&entrytime=656",
          "url": "https://pasadena.civic.fm/episodes/2026-10-30/highlights/1"
        }
      ],
      "episode": {
        "title": "Drone training grant, child mental health funds, and a work/live rethink",
        "url": "https://pasadena.civic.fm/episodes/2026-10-30",
        "item_url": "https://pasadena.civic.fm/episodes/2026-10-30?item=m13-2",
        "published_at": "2026-10-30T15:37:03Z",
        "audio": "https://op3.dev/e,pg=43cd7295-7008-5288-a5af-e4f041a7c1e5/civicfm-demo-media.t3.tigrisfiles.io/pasadena/episodes/2026-10-30-2bf34825.mp3",
        "covered": true
      },
      "updated_at": "2026-10-30T15:37:03Z",
      "comment": {
        "email": "correspondence@cityofpasadena.net",
        "form": null
      }
    }
  ],
  "total": 3,
  "next": "MQ=="
}

That's cut short for the page: the API gives this item's other 7 points too.

More to try:

• Everything under a topic: /api/v1/items?topic=surveillance
• One show's items since a date: /api/v1/items?show=pasadena&since=2026-10-27
• Every item with a motion that failed: /api/v1/items?result=failed
• This item alone, by its id: /api/v1/items/pasadena:13:2
• What's changed since a time: /api/v1/items?updated_since=2026-10-30T15:37:03Z
• Two at a time, to see a next: /api/v1/items?limit=2

Page by page

A response holds 50 items, or as many as limit asks for, up to 200. When more match, next is a cursor: send it back as cursor, with the same filters, for the page after. When next is null, that was the last page.

curl 'https://civic.fm/api/v1/items?limit=2'
curl 'https://civic.fm/api/v1/items?limit=2&cursor=Mg=='

Keeping a copy

Ask for what's changed since the latest updated_at you hold. It comes oldest change first, so you can stop anywhere and go on from the last one you read:

curl 'https://civic.fm/api/v1/items?updated_since=2026-10-30T15:37:03Z'

An item's id never changes, so a changed item replaces yours. An episode that's retracted takes its items with it: GET /api/v1/items/{id} then answers 404, which is how a copy learns of it.

When it can't answer

A mistake is answered in JSON too, {"error": …}, saying what to ask instead: 400 for a filter it can't read, or one it needs and didn't get, 404 for a path, an item, or a show it doesn't have, and 405 for anything but GET or HEAD. A parameter it doesn't know is ignored.

So /api/v1/items?topic=drones answers 400:

{"error": "There's no topic 'drones': /api/v1/topics lists them."}

From a web page

A script on any site can call it:

const r = await fetch('https://civic.fm/api/v1/items?topic=surveillance')
const { items, total, next } = await r.json()

Every response says so, and how long it may be kept:

Access-Control-Allow-Origin: *
Cache-Control: public, max-age=300

Endpoints

GET /api/v1/items

Items, newest meeting first, or best match first when you give q, or oldest change first when you give updated_since. Each filter is optional.

qstring	Words to look for, in the item's title, what it is, its votes, and what was said. "Quotes" match a phrase, and -word leaves one out.
topicstring	A topic's slug, from /api/v1/topics.
showstring	A show's slug, from /api/v1/shows.
sincestring (date)	The earliest meeting date to give, YYYY-MM-DD.
untilstring (date)	The latest meeting date to give, YYYY-MM-DD.
resultstring	The items with a motion of that result: passed, failed, continued, withdrawn, no_action, or unclear.
updated_sincestring (date-time)	A time, like 2026-10-01T13:00:00Z: the items whose updated_at is at or after it, oldest change first. A time that names no zone is UTC.
cursorstring	The next a response gave, for the page after it.
limitinteger	Items per page: 50 if you don't say, and 200 at most.

It answers:

itemsarray of items	The items on this page.
totalinteger	How many match in all.
nextstring or null	The cursor for the page after: send it back as cursor, with the same filters. null on the last page.

GET /api/v1/items/{id}

One item, by its id, as items has it. 404 if there's none, which is also how a copy learns that an episode was retracted.

GET /api/v1/topics

The topics, by area, each with how many items it has.

It answers:

areasarray of objects	Every area, each with its topics, whether or not they have items yet.
areas[].slugstring	Its name in addresses.
areas[].labelstring	What it's called.
areas[].topicsarray of objects	Its topics.
areas[].topics[].slugstring	Its name in addresses, and what the topic filter takes.
areas[].topics[].labelstring	What it's called.
areas[].topics[].aboutstring	What belongs in it.
areas[].topics[].countinteger	How many items it has.

GET /api/v1/shows

The shows, each with the place it covers, its site, and its feed.

It answers:

showsarray of objects	Every public show, by title.
shows[].slugstring	Its name in addresses, and what the show filter takes.
shows[].titlestring	What it's called.
shows[].placestring	The place whose meetings it covers.
shows[].regionstring	Where that place is: its state.
shows[].statestring	Whether it's still publishing: live, paused, or retired.
shows[].sitestring (uri)	Its site.
shows[].feedstring (uri)	Its podcast feed.
shows[].appsobject	Where to subscribe to it: the podcast apps it's listed in.
shows[].apps.applestring (uri) or null	Its listing in Apple Podcasts, or null where it isn't listed yet.
shows[].apps.spotifystring (uri) or null	Its listing in Spotify, or null where it isn't listed yet.
shows[].apps.youtubestring (uri) or null	Its podcast on CivicFM's YouTube channel, each episode as a video, or null before its first is up.
shows[].digestboolean	True for a show of shows, made from its area's shows' episodes. Their items are listed under them, so it has none of its own.
shows[].bodiesarray of objects	The bodies whose meetings it covers. A digest has none.
shows[].bodies[].idstring	Its name in the show's config.
shows[].bodies[].labelstring	What it's called, as an item's meeting.body has it.
shows[].bodies[].agendasstring (uri) or null	The government's own page of its meetings ahead and their agendas, or null where CivicFM doesn't have it. CivicFM reports a meeting after it happens: this is where to see what's coming up.
shows[].coversobject or null	What it has published so far, or null before its first episode.
shows[].covers.firststring (date) or null	The earliest meeting date it has items from. null for a digest, whose items are its members'.
shows[].covers.laststring (date) or null	The latest meeting date it has items from: nothing after it has been reported yet. null for a digest.
shows[].covers.episodestring (date-time)	When its latest episode went out, UTC.

GET /api/v1/places

A city, town, county, or ZIP code: the governments over it, each with the show that covers it, and the shows to listen to for it, roundups included. Words that fit several places answer only choices.

qstring

A city, town, or county, with its state if you know it (Anytown, VT), or a ZIP code, which names one place.

It answers:

placestring or null	The place as the Census Bureau names it, or the ZIP code. null when the words fit several places: choices lists them.
kindstring or null	What it is: a city, town, county, … as governments has them, unincorporated for a place no city or town governs, or zip.
governmentsarray of objects	The governments over it: its own, then its county's. An unincorporated place has only its county's, and a ZIP code each one it falls in. No school or special districts.
governments[].namestring	Its name, as the Census Bureau's list has it.
governments[].kindstring	What that list calls it: city, town, village, borough, township, county, or community.
governments[].showstring or null	The slug of the show that covers it, as show takes it, or null where no show covers it yet.
notestring or null	What a person there should know of who governs them, where it's not plain: that a place is unincorporated, so its county governs it. Or null.
listenarray of objects	The shows to hear for it: those that cover its governments, then the roundups that carry them. Empty where no show covers it yet.
listen[].slugstring	Its name in addresses, and what show takes.
listen[].titlestring	What it's called.
listen[].sitestring (uri)	Its site, with each episode.
listen[].feedstring (uri)	Its podcast feed.
listen[].appsobject	Where to subscribe to it: the podcast apps it's listed in.
listen[].apps.applestring (uri) or null	Its listing in Apple Podcasts, or null where it isn't listed yet.
listen[].apps.spotifystring (uri) or null	Its listing in Spotify, or null where it isn't listed yet.
listen[].apps.youtubestring (uri) or null	Its podcast on CivicFM's YouTube channel, each episode as a video, or null before its first is up.
listen[].digestboolean	True for a roundup of several shows' weeks, which carries the place's own.
choicesarray of strings	When the words fit several places, up to five of them, as place would name each: ask again with one. Otherwise empty.

GET /api/v1/briefing

What a place's local governments did lately: the stories their shows' latest episodes told in the last 14 days, as the front page ranks them, with those episodes. Give place or show.

placestring	A city, town, or county, with its state if you know it (Anytown, VT), or a ZIP code, which names one place.
showstring	A show's slug, from /api/v1/shows, in place of place. A roundup's is its shows'.

It answers:

placestring or null	The place briefed, as the Census Bureau names it, or null when a show was asked for.
storiesarray of items	The stories, eight at most, from the shows' latest episodes out in the last 14 days: a statewide roundup's top stories that are theirs, then an area roundup's, then each show's own, its first before its second. Empty with nothing out in 14 days.
episodesarray of objects	Those episodes, to listen to: each show's own, then the roundups'.
episodes[].showobject	Its show.
episodes[].show.slugstring	Its name in addresses, and what the show filter takes.
episodes[].show.titlestring	What it's called.
episodes[].show.sitestring (uri)	Its site.
episodes[].titlestring	What it's called.
episodes[].published_atstring (date-time)	When it went out, UTC.
episodes[].urlstring (uri)	Its page, with the player.
episodes[].audiostring (uri) or null	Its MP3, as its show's feed gives it.
episodes[].digestboolean	True for a roundup of several shows' weeks.
searchedarray of objects	Each show briefed, with what it covers. Nothing after a show's covers.last has been reported yet.
searched[].slugstring	Its name, as show takes it.
searched[].titlestring	What it's called.
searched[].coversobject or null	What it has published so far, or null before its first episode.
searched[].covers.firststring (date) or null	The earliest meeting date it has items from. null for a digest, whose items are its members'.
searched[].covers.laststring (date) or null	The latest meeting date it has items from: nothing after it has been reported yet. null for a digest.
searched[].covers.episodestring (date-time)	When its latest episode went out, UTC.

GET /api/v1/shows/{slug}/officials

A show's bodies: where each takes written comment, and who holds each of its seats today, with their office's email, phone, and page where CivicFM has them. A committee drawn from another body's seats names that body.

It answers:

showobject	The show.
show.slugstring	Its name in addresses, and what the show filter takes.
show.titlestring	What it's called.
show.sitestring (uri)	Its site.
todaystring (date)	The day the seats are read for, where the show is, YYYY-MM-DD.
bodiesarray of objects	Each body whose meetings it covers.
bodies[].idstring	Its name in the show's config.
bodies[].labelstring	What it's called, as an item's meeting.body has it.
bodies[].commentobject or null	Where it takes written comment, as its agenda says, or null where CivicFM doesn't have it.
bodies[].comment.emailstring or null	The address its agenda gives for written comment, or null.
bodies[].comment.formstring (uri) or null	The page its agenda gives to send written comment from, or null.
bodies[].agendasstring (uri) or null	The government's own page of its meetings ahead and their agendas, or null where CivicFM doesn't have it. CivicFM reports a meeting after it happens: this is where to see what's coming up.
bodies[].membersarray of objects	Who holds each of its seats today, as the show's roster has them. Staff, who hold no seat, aren't listed. Empty for a committee drawn from another body's seats, and where the roster lists no one.
bodies[].members[].namestring	Their name.
bodies[].members[].titlestring or null	The titles they hold today.
bodies[].members[].contactobject or null	Their office's contact, or null where CivicFM doesn't have it.
bodies[].members[].contact.emailstring or null	The office's email, or null.
bodies[].members[].contact.phonestring or null	The office's phone, as the government writes it, or null.
bodies[].members[].contact.pagestring (uri) or null	The office's page on the government's site, or null.
bodies[].drawn_fromstring or null	For a committee whose members are some of another body's, that body's label. Otherwise null.

GET /api/v1

Where each of these is, with the page for agents, the reference, and the description.

It answers:

docsstring (uri)	The page for agents: the MCP server, then the API.
itemsstring (uri)	The items.
topicsstring (uri)	The topics.
showsstring (uri)	The shows.
referencestring (uri)	The API's reference, to try each call in.
openapistring (uri)	The API's description, as OpenAPI.
mcpstring (uri)	The MCP server, for an AI assistant to read the same items: add it as a connector.
placesstring (uri)	A place's governments, and the shows to hear for it.
briefingstring (uri)	What a place's governments did lately.

An item

One thing a body took up at a meeting: what it is, how it was voted on, what was said about it, and where each of those is in the recording.

An outcome

One motion and what became of it: a vote, as data. Its fields say what its text says in a sentence. On an outcome that went out before the API had them they're null, if its meeting's notes have changed since. Its text is always there.

A point

Something an official or staff said, paraphrased and never quoted.

A person

Someone who voted or spoke, named as the episode's page names them: an official only when the recording shows who it was, and a member of the public never. So a person is {"name": "Ana Lee", "role": "Councilmember, District 5"} when the recording shows who, and {"name": null, "role": "a councilmember"} when it doesn't.