On this page Connect What your agent can ask Signing in Products How Starlight bills The same answers as files

Connect your agent to Starlight

The Starlight MCP server gives any AI agent the current Starlight products and every price, straight from the list that calculates each charge. Products and prices are free to read and need no sign-in. When you sign in to your Starlight account, an agent can also see your balance, your Sluice recordings and what each has cost, and your Cadence tasks and plan. If you also allow it, the agent can manage your Cadence tasks for you, and stop a Sluice recording or change the words it looks for. It only ever gets what you allow.

Server address

https://mcp.starlightglobal.industries/mcp

Connect

In any app or agent that supports remote MCP servers, add a server with the address above. Products and prices work straight away, with no sign-in or key.

In Claude, open Settings, then Connectors, choose Add custom connector, and paste the address above.

With Claude Code, run:

claude mcp add --transport http starlight https://mcp.starlightglobal.industries/mcp

What your agent can ask

Without signing in

  • List Starlight products list_products
    Every Starlight product: what it does, where it is, and whether it is paid for from Starlight credits.
  • Get Starlight prices get_prices
    The current price list for a Starlight product, or for every product when none is named. Each price is exact, some are fractions of a cent, and the list carries the version every statement line records.
  • How Starlight bills how_starlight_bills
    How Starlight counts and prices work: small, whole pieces of work, each priced from what it costs, paid from one balance of Starlight credits.
  • Estimate a Sluice recording estimate_recording
    What a Sluice recording of a given length is expected to cost and how long it is expected to take, before you record it. The cost is a range from the published price list, at a typical number of things in view: the whole recording at the cheapest off-peak price, and at the peak price, which is the most it can cost. The time is a planning range for processing, with the slowest figure measured so far beside it, and the upload time when you give your upload speed. After you sign in, the recording's own cost and timeline say what it actually took.
  • A link to upload a recording to Sluice upload_recording_link
    A link that opens Sluice's upload page with the project, the words to look for and the transcription choice already filled in, so the person can choose their recordings in their browser and upload them there. The files go straight from the person's browser to Sluice and never pass through the chat. Nothing is uploaded or charged until the person opens the link, signs in, chooses the files and starts the upload, and Sluice shows the cost first. Words are the things to look for, such as "cordless drill" or "paint cans". Needs no sign-in.

After you sign in

  • Estimate what Sluice would charge estimate_cost
    What Sluice would charge to process a recording, worked out from the published price list before anything is uploaded: the most it can cost, at the peak price, with a breakdown by step and unit, and the least, at the cheapest off-peak price. Give its length in seconds and, to include receiving and holding the file, its size in bytes. A real recording is charged for the outlines found in it, so give masks_per_frame for a busier or sparser scene than the usual 15. Each prompt finds at most 200 outlines in a frame and a frame can be prompted any number of times, so a busy frame can hold thousands; give prompts to check masks_per_frame against that. The charge is per outline, not per prompt. Needs sign-in.
  • Your Starlight balance get_account
    Your Starlight credits balance, how much of it is complimentary credit and when that expires, whether your account is in good standing, and whether new work can start. Needs sign-in with permission to see your balance.
  • Your Sluice recordings list_recordings
    Every one of your Sluice recordings, newest activity first, including ones still uploading or not yet charged: its state, how much of it has uploaded, what it has cost so far, and the details that set each apart: the name, size and length of each file it was made from, which is the file to choose again to continue an upload that stopped, its length, resolution, frame rate, codec, bitrate, size and files, and how many frames, outlines and items processing produced. A detail is null where it was not recorded, and older recordings carry only their length and frame rate. When there are more, the answer gives next_before: pass it back as before for the next page. Needs sign-in with permission to see your Sluice recordings.
  • One of your Sluice recordings get_recording
    Where one of your Sluice recordings has got to: its state, how many of its passes have finished and what share of the frames they cover, how far finding, describing and grouping objects have each got, the transcript's word count, each piece of work with its attempts and the error or reason it failed or waits, the items found so far and when each appears, how sure the model was that each thing it looked for was in view at its best frame, and every problem noted with its detail. It also carries what the recording's page shows besides: the options it was opened with, its files, whether it is held and why, what it can do now (stop, continue, move to a rented GPU), the processing time used against its limit, its notices, and what is held for it. Each pass looks at twice the frames of the one before, so share_of_frames, not the count of passes, is how much of the work is done. Passes are counted from 1. While its passes run it carries eta: when they are estimated to finish (finish_at, with earliest_at and latest_at), measured from the time per frame of its own finished passes, and the basis for it; known is false, with the reason, until enough passes have finished to measure. Each queued piece of work is listed under waits: when it is held back until a set time, that time and why; on the graphics card, how many steps are ahead of it, how many are running, and an estimated start (starts_at) from that machine's recent step times, with no time given when a running step has run past its usual time. It carries the first 200 items, and says how many there are in all. Needs sign-in with permission to see your Sluice recordings.
  • What a Sluice recording cost get_recording_costs
    What one of your Sluice recordings has cost so far: the total, the total for each stage and each unit, and how much of it is on an invoice and how much is on your balance still to be invoiced, with the recording's details: its length, resolution, frame rate, codec, bitrate, size and files. Every charged_micro is money charged, in millionths of a dollar, and each list sums to the total. Needs sign-in with permission to see your Sluice recordings.
  • Where a Sluice recording's time went get_recording_timeline
    Where one of your Sluice recordings spent its time: when each piece of work started and finished, how its time split between the graphics processor, the processor, decoding, waiting for the graphics processor and reading or writing, how much of the job's wall-clock time something was running and how much it waited, the longest waits and what they fell between, the time per stage, and the job's timeline marks. The recording's details (length, resolution, frame rate, codec, bitrate, size, frames examined, items) come with it, so the time can be read against what was processed. Every *_ms is milliseconds. A started_at is the finish less the compute time, so it is an estimate. Needs sign-in with permission to see your Sluice recordings.
  • Which Sluice items have a picture get_recording_pictures
    How many of the items found in one of your Sluice recordings have a picture to show, and why the rest do not: the picture is still on the processing computer, it has not been reported yet, or the item has no picture at all. Also lists which passes over the recording have found items and which have made sharper pictures. Counts only, never the pictures. Needs sign-in with permission to see your Sluice recordings.
  • The words a Sluice recording looks for get_recording_words
    The words one of your Sluice recordings looks for, and the ones it does not: each word's id, its text, where it came from (typed, the project's earlier words, or heard in the narration), whether it is kept, proposed or removed, whether it is looked for in the whole recording or only around the times it was said, when it was said, how many items it found, and whether the transcriber was unsure of it. Also says whether the words wait for a review (status waiting, which holds the closer passes until someone confirms or 15 minutes pass), and the version a change to the words is made against. Read only. Needs sign-in with permission to see your Sluice recordings.
  • Change the words a Sluice recording looks for change_recording_words
    Keep, remove, rename or add the words one of your Sluice recordings looks for, against the version of the words it was read at. A change with an id changes that word: its text, its state (kept or removed) or where it is looked for (whole, or windows around the times it was said). A change without an id adds a typed word, looked for in the whole recording. confirm true ends the review, so the closer passes start now with these words instead of after the 15-minute wait; it starts no work the recording would not do anyway. save_to_project true makes the kept whole-recording words the project's words for its next recording. When the words changed since that version, nothing is changed and the answer gives the current words and version. Changes only the words: never a charge, the recording itself or another recording. Needs sign-in with permission to change your Sluice recordings.
  • Stop a Sluice recording stop_recording
    Stop one of your Sluice recordings that is processing: its queued work is held and the step running now finishes first, so no new work starts. The member continues it from the recording's page in Sluice; this tool cannot continue it, because continuing spends credits. Stopping a recording that is already stopped answers its state again. Needs sign-in with permission to change your Sluice recordings.
  • Rotate an Iris item's picture rotate_iris_picture
    Rotate one of your item pictures as Iris and Sluice show it: 90, 180 or 270 degrees clockwise, or 0 to show it as it was taken. The rotation is from the picture as it was taken, which is how the Iris item answers here show it, so look at the picture first and pass the rotation that makes it upright. Rotating it again replaces the last rotation. The stored picture is never changed, so its evidence and its links stay valid. Pass the item's picture link from an Iris item answer. Needs sign-in with permission to change your Sluice recordings.
  • What a Sluice recording is missing get_recording_gaps
    What one of your Sluice recordings could show and does not yet, such as step-by-step playback or where an item sits in its frame: each gap's code, what is missing, what the recording's page says about it in the same words, whether Sluice is already adding it, and whether a free fix is offered. The codes are the ones the page shows, so this answers why an item has no frame the way the page does. Up to 32 gaps; the answer says how many were left out. Read only: it fixes nothing, and each free fix is offered on the recording's page in Sluice. Needs sign-in with permission to see your Sluice recordings.
  • A report over your Sluice recordings build_report
    The facts of a report over 1 to 100 of your Sluice recordings: for each recording, its page, the source file you name, its sha256, its length, how deep it was read and what it cost line by line; for each item found in them, its canonical record, its Iris page, the evidence for each field and when in the source file it was seen. Name each recording by its job id and the full address of its source file as you hold it, such as a path on your NAS; Sluice never works the address out itself, and an address that expires, such as a signed link, is refused. Each recording must be one finished file: one that has not been read yet, or was uploaded in parts, is refused with the reason. The same recordings always give the same report_id, and digest changes only when the facts do: Sluice keeps each new revision, so asking again with nothing changed gives the same revision, and the answer links to the report on Iris, where the member can review it and download it. Every item carries its whole record, so an answer over many recordings is long. Building a report is not charged, but each kept revision is charged as storage by the day, at the same rate as the evidence it describes. Needs sign-in with permission to see your Sluice recordings.
  • A report Sluice kept get_report
    A report built earlier over your Sluice recordings, read back by its report_id: the facts of its last revision, or of the revision you name, the notes added to it, and every revision number Sluice kept. The facts have the same shape as a newly built report, and the answer links to the report on Iris. Read only. Needs sign-in with permission to see your Sluice recordings.
  • Your Iris items list_iris_items
    The items Sluice found in your recordings, as Iris shows them: each item's job and item id, the word that found it and every other word that did, its project, when in the recording it was seen and for how long, how many views it has, how clear its picture is, when it was recorded, a link to it in Iris, and a link to its picture. A picture link is signed and works without sign-in for about a day, so treat it like the picture itself; it is null when the picture is not held yet. Up to 50 items a call: when there are more, the answer gives next_offset, which you pass back as offset. Most recently recorded first unless sort says otherwise; the order is fixed, so pages never repeat or skip an item. Each item also says what found it (each word with how many outlines it found), the recording it was in with a link to it in Sluice, when the camera recorded it (or that this is not known), when it was uploaded, and how many days before this answer it was recorded or uploaded (seen.age_days, with seen.age_basis saying which), when in the recording it was seen and the moment its picture was taken, and where it is: in a named container, no place recorded, or not known. The answer says how many of your recordings were searched and how many are not read yet or still being processed, so an empty answer is not a certain no. It does not search your reviewed inventory. Also gives the words and projects under the same filters, each with how many items it has, so you can narrow by one. Read only. Needs sign-in with permission to see your Iris items.
  • One of your Iris items get_iris_item
    One of your Iris items by its job and item id: the same fields the Iris item list gives each item, with a link to it in Iris and a signed link to its picture that works without sign-in for about a day. Each part of the answer is on its own line with what it rests on: the words that found it, the recording, when the camera recorded it and when it was uploaded, how many days before this answer that was, when it was seen, the picture's moment and its evidence id, where it is, and the evidence ids for each field. A part with no evidence says not known. The older the sighting, the less likely the item is still where it was seen. The picture itself comes back in the answer as an image, so you can look at it, when it is held, no larger than 1 MB and no more than 2000 pixels on a side; otherwise the answer says why it is not there. Read only. Needs sign-in with permission to see your Iris items.
  • Look at your Iris items' pictures view_iris_pictures
    The pictures of your Iris items, as images you can look at, a few at a time: the same items, filters and order as the Iris item list, each picture after a line naming its item. Up to 10 items a call (5 unless limit says otherwise): when there are more, the answer gives next_offset, which you pass back as offset. A picture is left out, and the answer says why, when it is not held yet, is larger than 1 MB, is more than 2000 pixels on a side, or would take the answer past 5 MB of pictures. Read only. Needs sign-in with permission to see your Iris items.
  • Iris items that look like one of yours find_similar_iris_items
    More like this: your Iris items that look most like one you name, across every project you hold or the one you name, closest first, each with a similarity from 0 to 1 and the same fields the Iris item list gives. It compares how the items look, not their words, so it finds the same or a similar thing across recordings. The item itself is never in the answer. Up to 48 a page and the closest 120 in all: when there are more, the answer gives next_offset, which you pass back as offset. An item with no picture data yet answers embedded false and no items. Read only. Needs sign-in with permission to see your Iris items.
  • Find words in your Iris items find_iris_words
    The words your Iris items were found by that match some typed text, each with how many items it found: the exact word first, then its singular, then words that start with it, then the rest, and within each, most items first and then alphabetically. Typed drills finds drill, then drill bit, then power drill. Each word can be used to list the items it found. Your reviewed inventory is not searched here. The answer says how many of your recordings were searched, so no match is not a certain no. When there are more, the answer gives next_offset: pass it back as offset for the next page. Read only. Needs sign-in with permission to see your Iris items.
  • Your reviewed inventory list_inventory_items
    Your reviewed inventory items: the ones imported from Homebox and the ones Streamline passed, which a search of your recordings does not cover. Each item gives its title, its id, where it is (the location recorded in the inventory, or no place recorded), the container it is in, who reviewed it and when, where it came from, and the evidence ids for each field. q matches the item's text, such as cordless drill; leave it out for every item. Up to 50 items a call: when there are more, the answer gives next_offset, which you pass back as offset. The answer says how many of your inventory workspaces were read and how many could not be, so an empty answer is not a certain no. Read only. Needs sign-in with permission to see your Iris items.
  • One of your reviewed inventory items get_inventory_item
    One of your reviewed inventory items by its id: its title, where it is and on what basis, the container it is in, its tags and categories, who reviewed it and when, where it came from, links to its pictures and files, and the evidence ids for each field. A part with no evidence says not known. Read only. Needs sign-in with permission to see your Iris items.
  • What each of your Streamline projects is for get_streamline_goals
    Your Streamline workspaces and the goal an owner set for each Sluice project in them: sell its items, keep a record of them, insure them or make a claim, or clear them out. Each project says how many items it holds, whether the queues take it, and when its goal was set; a project with no goal says so. A workspace whose data processing terms you have not accepted is listed without its projects. Up to 10 workspaces a call. Read only. Needs sign-in with permission to see your Streamline goals.
  • The questions waiting in one of your Streamline workspaces list_streamline_questions
    The questions Streamline is waiting on a person to answer in one workspace, highest value first: possible duplicates, items to split, items to identify, price or place. Each says what it asks about, the items it concerns by id and title, and the value at stake when one is known. It answers nothing: the answers are given in Streamline, whose lane page the answer links. Up to 25 questions a call. Read only. Needs sign-in with permission to see your Streamline goals.
  • The next steps toward your goals in one Streamline workspace get_streamline_plan
    The plan for one Streamline workspace: the next steps toward what the member wants to do with each item, grouped by room. Each step says what to do, how many items it concerns, the goal it serves, the member's own words for that goal when they gave them, and whether the member takes it or Streamline is already doing it. A step that asks an owner to decide a proposed goal is decided in Streamline, whose plan page the answer links. The answer says when the plan was made from only part of the workspace. Up to 50 steps a call. Read only. Needs sign-in with permission to see your Streamline goals.
  • What happened lately in one Streamline workspace get_streamline_activity
    The recent activity of one Streamline workspace, newest first: items merged, kept apart or given details, questions answered, goals stated, confirmed or declined, each time Streamline looked for goals to suggest, and what was undone, with when and whether a member, an assistant or Streamline did it. Read it to see what changed since a goal was set. Undo is done in Streamline, whose activity page the answer links. Up to 50 lines a call. Read only. Needs sign-in with permission to see your Streamline goals.
  • What was decided about your goals in one Streamline workspace get_streamline_goal_history
    The goal history of one Streamline workspace, newest first: every goal stated for the workspace, a project, a room or an item, in the member's own words when they gave them, whether a member or an assistant stated it, and whether it was confirmed, declined, withdrawn or is still waiting on a member. Read it before suggesting a goal: a goal a member declined for a scope should not be suggested there again. Goals are confirmed or declined in Streamline, whose plan page the answer links. Up to 100 statements a call. Read only. Needs sign-in with permission to see your Streamline goals.
  • The items in one of your Streamline workspaces list_streamline_items
    The live items in one Streamline workspace, by title: what each is, how far Streamline has identified it, the room it is kept in, whether it is ready or waiting on a person, its value when known and the open question it is in. Narrow it to one Sluice project, to ready, needs_you, unpriced, scenes or collated items, or search titles, kinds and rooms by words. Pictures are left out. Up to 50 items a call; pass the answer's next value to read the page after. Read only. Needs sign-in with permission to see your Streamline goals.
  • One item in a Streamline workspace get_streamline_item
    One item in a Streamline workspace: its title, kind and room, what it is identified as at each level (kind, type, brand, model, variant, unit) and how sure Streamline is of each, its condition, its value and, once valued, what similar items sold and are asking for, the recording it was seen in and when in it, the words heard about it there, the open question it is in, and the goal that applies to it: the item's own, or the one it takes from its room, project or workspace, or that nobody has said yet. Pictures are left out. Read only. Needs sign-in with permission to see your Streamline goals.
  • Name one of your Streamline items with Claude label_streamline_item
    Ask Claude to name and price one of your Streamline items from its chosen pictures: a title, how sure it is of the kind, type, brand, model, variant and condition, the markings it read, an eBay listing (category, item specifics, description) and a suggested price with the sold and asking listings it found by searching the web. It proposes a label and changes nothing in Streamline. Each call is paid from your Starlight credits at the Streamline label price, within the spending limits you set when you allowed this app; the answer says what it cost. The workspace must be paid from your own credits, and its owner has accepted Anthropic's terms for it in Streamline. Needs sign-in with permission to label your Streamline items.
  • Your Scan items list_scans
    The things you captured with Scan, newest first: each item's name, description and tags, when it was captured and last changed, and the files it holds, each by kind and size. Items in Scan's Trash are left out. The pictures themselves stay in Scan: the answer links to the Scan app, where you see them signed in. Up to 50 items a call: when there are more, the answer gives next_before, which you pass back as before. Read only. Needs sign-in with permission to see your Scan items.
  • One of your Scan items get_scan
    One of your Scan items by its id: the same fields the Scan item list gives, and where it is stored when you placed it in a location. The pictures stay in Scan: the answer links to the Scan app. Read only. Needs sign-in with permission to see your Scan items.
  • Send feedback to Starlight send_feedback
    Sends feedback to Starlight Global Industries on behalf of the signed-in member: a problem they hit, something that did not work as they expected, or an idea. At most 500 characters. The report is filed with the member's account, which Starlight keeps only until the report is filed, and with the name of the app that sent it. Each account can send 5 reports an hour. Needs feedback:write, which the member allows on the Starlight sign-in page.
  • Starlight connection status connection_status
    Which Starlight account this connection signs in to, what its owner allowed, and until when. Needs sign-in.
  • Your Cadence plan get_plan
    Your Cadence calendar: the time blocks Cadence has planned for your tasks on the days you ask for, your events on those days, and every task it could not fit or that will finish after its deadline. The plan is worked out from your tasks, hours and settings each time, exactly as the Cadence app does, so it moves as time passes and as tasks change. A long plan comes in pages, and at most 50 events, 50 unfitted tasks and 50 late tasks are listed, each with its total. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Your Cadence tasks list_tasks
    Your Cadence tasks, soonest planned first: each one's project and stage, how much time is left, its priority and deadline, and when it is next planned or why it could not be. Occurrences of repeating tasks inside the planning horizon are included. When there are more, the answer gives next_offset: pass it back as offset for the next page. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • One Cadence task get_task
    Everything about one Cadence task: its fields and notes, every block it is planned in, and whether it is late or could not be placed and why. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Your Cadence projects list_projects
    Your Cadence projects, each with its stages in the order they must happen and the open tasks in each stage in their set order, with the time left in each. In a project with stages, no task starts before the tasks of earlier stages are planned to finish; in a sequential project each task also waits for the one before it. When there are more, the answer gives next_offset: pass it back as offset for the next page. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Your Cadence events list_events
    The fixed events in Cadence (appointments and other busy time no task is planned over) between two days, earliest first. When there are more, the answer gives next_offset: pass it back as offset for the next page. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Your Cadence repeating tasks list_repeating_tasks
    Your Cadence repeating tasks: how often each repeats, how long it takes, when each occurrence is due, and its next occurrences and whether they are done. When there are more, the answer gives next_offset: pass it back as offset for the next page. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Your Cadence hours and settings get_cadence_settings
    Your Cadence settings (time zone, how far ahead it plans, block size and breaks, reminder lead time) and your schedules, each a named set of weekly working hours that tasks are planned into. Needs sign-in with permission to see your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs.
  • Add or change a Cadence task save_task
    Adds a task to Cadence, or changes one when task_id is given; a change sets only the fields you give, and null clears a field. Cadence then plans it into your working hours; the answer says when, or why it could not. Changing one occurrence of a repeating task changes that occurrence alone. A date is YYYY-MM-DD and a time is YYYY-MM-DDTHH:MM, both in your Cadence time zone, or a full timestamp with its offset. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Change many Cadence tasks at once update_tasks
    Finds the tasks that match where and makes the same change to each: the fields in set, with null clearing a field. Every condition in where must hold, and at least one besides status is needed. It previews by default: the answer lists what would change, each task's values before and after, and how the plan would move, and nothing is saved until dry_run is false. At most 200 tasks change in one call; a larger match is refused whole. Occurrences of repeating tasks are not matched. A task moved to another project or stage goes last there, keeping the order it had. A date is YYYY-MM-DD and a time is YYYY-MM-DDTHH:MM, both in your Cadence time zone, or a full timestamp with its offset. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Mark a Cadence task done complete_task
    Marks a task done, or not done again with done false. For an occurrence of a repeating task (an id ending in a date) only that occurrence is marked. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Mark many Cadence tasks done at once complete_tasks
    Finds the tasks that match where and marks each one done, or not done again with done false. Every condition in where must hold, and at least one besides status is needed. status in where is open when left out, or done when done is false. It previews by default: the answer lists every task that would be marked and how the plan would move, and nothing is saved until dry_run is false. At most 200 tasks change in one call; a larger match is refused whole. Occurrences of repeating tasks are not matched. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Move a Cadence task move_task
    Moves a task to a project and stage, at a position in that stage's order. Order decides which of otherwise equal tasks is planned first, and in a sequential project which waits for which. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Add or change a Cadence project save_project
    Adds a project, or changes one when project_id is given. stages, when given, is the whole list of stages in the order they must happen: keep a stage by giving its stage_id (to rename or reorder it), add one by leaving stage_id out, and remove one by leaving it out of the list, which takes its tasks out of any stage. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Add a Cadence project with its tasks create_project_plan
    Adds a whole project at once: its stages in order and the tasks in each, in order. At most 199 tasks. The answer says how much time it takes and when Cadence plans it to finish. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Add or change a Cadence event save_event
    Adds a fixed event (busy time no task is planned over), or changes one when event_id is given. A date is YYYY-MM-DD and a time is YYYY-MM-DDTHH:MM, both in your Cadence time zone, or a full timestamp with its offset. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Add or change a Cadence repeating task save_repeating_task
    Adds a task that repeats, or changes one when repeating_task_id is given. freq is daily, weekly or monthly, every interval days, weeks or months from starting_on. A weekly one repeats on by_day; a monthly one on day_of_month, or on the nth by_day of the month when nth is given (1 to 5, or -1 for the last). Each occurrence is due due_days after its date. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Add or change Cadence working hours save_schedule
    Adds a schedule (a named set of weekly working hours tasks are planned into), or changes one when schedule_id is given. windows, when given, replaces every day's hours. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Change Cadence settings update_cadence_settings
    Changes your Cadence settings; only the fields you give change. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Delete something from Cadence delete_cadence_item
    Deletes a task, project, event, repeating task or schedule. Deleting a project keeps its tasks, without a project. A schedule that tasks still use cannot be deleted, nor your last one. To skip one occurrence of a repeating task, mark it done instead. A deletion cannot be undone from here: Cadence keeps only the id of what was deleted. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.
  • Delete many Cadence tasks at once delete_tasks
    Finds the tasks that match where and deletes them. Every condition in where must hold, and at least one besides status is needed. It previews by default: the answer lists every task that would be deleted and how the plan would move, and nothing is deleted until dry_run is false. At most 200 tasks are deleted in one call; a larger match is refused whole. Occurrences of repeating tasks are not matched. A deletion cannot be undone from here: Cadence keeps only the id of what was deleted. Needs sign-in with permission to change your Cadence tasks, and AI assistant access turned on in Cadence (Account, AI assistants) on a device that syncs. The change reaches your devices the next time each one syncs.

Signing in

The first time your agent asks about your own account, your app opens a Starlight sign-in page. Sign in to your Starlight account, then choose what the app may see and for how long: 30, 90 or 365 days. You can untick anything you would rather not share. Permission to change your Cadence tasks is a separate choice, and the sign-in page asks for it the first time your agent tries to make a change.

Four permissions let an app do more than read. Permission to change your Sluice recordings lets an app stop a recording that is processing and change the words a recording looks for, but not continue one, which spends credits. Permission to change your Cadence tasks lets an app add, change, complete and delete your Cadence tasks, projects, events, repeating tasks and schedules, and change your Cadence settings. It also needs AI assistant access turned on in Cadence, under Account, AI assistants. Permission to label your Streamline items lets an app ask Claude to name an item from its pictures, paid from your Starlight credits within the daily and monthly limits you set when you allow it. Permission to send feedback lets an app send your feedback to Starlight for you, up to 500 characters at a time and 5 reports an hour, and spends nothing. No permission lets an app change anything else in your Starlight account, or upload or delete a Sluice recording.

To disconnect an app, open your Starlight account and choose Disconnect under Connected agents. From then on it can no longer read your account.

Products

  • Sluice
    Record your surroundings. Sluice finds each item in your recordings and shows you the frames it came from, so you can check each one.
  • Iris
    Iris shows each item found in your Sluice recordings, with every picture of it and when it was seen.
  • Streamline
    Streamline is the app where your organization reviews the items Sluice finds. It settles what it can on its own, such as merging the same item found in different recordings and proposing names and values. It asks a person to decide only when it needs one. Today Streamline is open to invited organizations in early access, with a workspace for each organization that only its members can open.
  • Exchange
    Exchange shows what you own, an estimate of what it is worth today, and every option for it: keep, sell, donate, discard or claim. It is opening to people a few at a time, from a waitlist on its site.
  • Scan
    Scan brings photographs of your items into Starlight. Take a picture with your phone and Scan delivers it, even if you had no connection when you took it.
  • Cadence
    Cadence places your tasks into the hours you set aside for them and reminds you when a block is about to start. It works on your device without an account. With Starlight credit, it keeps an encrypted copy of your tasks so your other devices can open them.

How Starlight bills

One balance of Starlight credits pays for Sluice, Scan, Streamline and Cadence. Work is counted in small, whole pieces, each with a published price that starts from what the work costs us, plus a margin. Many prices are a tiny fraction of a cent, and every piece is counted at its exact price.

Work is counted in small, whole pieces

Starlight measures what it does for you one piece at a time: an item outlined in a frame, a picture cut out, an item named, a second of storage. Each piece has its own price, and your statement lists how many of each you used. You pay for those pieces, never for time on a graphics processor. A piece done on our own graphics processors costs less off-peak, when electricity is cheaper, and a piece done on a graphics processor we rent is charged the peak price.

Prices start from what the work costs us

Each price is what it costs us to do that piece of work, plus a margin. The margin also pays for the servers and services that keep Starlight running, so your statement lists only the work you asked for.

Tiny prices, counted exactly

Many pieces of work cost a tiny fraction of a cent, and some prices are measured in trillionths of a dollar, such as a megabyte stored for one second. Starlight counts every piece at its exact price, so your balance moves by what you actually used.

One balance that stays yours

The Starlight credits you buy pay for Sluice, Scan, Streamline and Cadence from the same balance, and they stay on your account until you spend them. If we cannot deliver what you paid for, we refund it.

The same answers as files

Agents and people that do not use MCP can read the same catalog directly.

Building on Starlight? The developer site explains how to connect an agent, what each tool does and how signing in works.