The server address is https://app.loopster.ai/api/mcp.

The AI tool signs in through OAuth, or it sends an API key in the header Authorization: Bearer loopster_sk_….

You make a key in the app. Open Settings, then Loopster API. See Create and revoke a Loopster API key.

With an API key, the tools work in the one workspace that owns the key.

A key has a role: Member — read only, Editor — read, write and run or Admin — everything an API key can do. Some tools need an Editor or Admin key. See "Which tools need more than a Member key?" below.

list_tables: lists the tables in your workspace, newest first.

get_table: shows one table with its columns and its settings, including Auto-run (the tools call it autoplay).

list_workbooks: lists your workbooks. Every table is in exactly one workbook.

list_folders: lists your folders. A folder holds workbooks.

list_columns: lists a table's columns and how each one is set up.

list_views: lists a table's saved views. An empty list is normal.

list_contexts: lists the ICPs, company descriptions and tier lists that columns read. It shows titles and kinds, not the text.

create_table: makes a table. Blank, webhook and search tables are free. Any other starter type runs its source while the table is created, and spends from your wallet.

rename_table: changes a table's title.

delete_table: moves a table to Trash. It does not stop work that is already running, so call stop_run first.

move_table: files a table into a workbook.

create_workbook: makes an empty workbook, at the top level or in a folder.

move_workbook: moves a workbook and its tables into a folder.

create_folder: makes an empty folder.

set_table_autoplay: turns the table's Auto-run on or off. Switching it on does not run the rows already in the table.

set_table_auto_dedupe: skips or deletes rows whose value in one column matches another row's.

set_table_create_columns_from_keys: lets a new key in inserted rows become a new column. It is off by default.

list_column_templates: lists the templates you can create columns from. You can filter by type, provider or tab.

get_column_template: shows one template with its settings and defaults.

create_column: makes a column from a template. An AI column made from a model template has no prompt, and it does not run until you set one.

create_columns: makes up to 20 columns in one call. It is all or nothing.

update_column: changes a column's header or setup. A column's setup decides what it costs, so set it before you run it.

delete_column: deletes a column and all its cells.

create_view and update_view: save how a table is displayed, with its own filters, sort order and visible columns. update_view replaces the filters, sorting or visible columns you send. It does not merge them.

delete_view: deletes a saved view.

insert_rows: adds rows. One call carries at most 25,000 rows. If the table's Auto-run is on, inserting also starts enrichment and spends from your wallet.

list_rows: reads rows a page at a time, with each cell's status and any error.

get_row: reads one row in full, with every cell's value and status, and the error message if a cell failed.

update_row: writes values into one row. Its description says that on a table with auto-rerun columns, writing a cell also re-queues the columns that read it, which spends from your wallet.

update_cells: writes many cells across rows in one call. On a table with auto-rerun columns, it re-queues the columns that read the changed cells in the same way as update_row, which spends from your wallet.

delete_rows: deletes rows by id. There is no "delete every row" form.

create_upload_link: gives you an address to upload one CSV or JSON-lines file.

import_file: imports an uploaded file into a table and returns a jobId. It starts no enrichment and costs nothing.

get_job: checks an import. Call it until status is complete or failed.

export_table: exports a table, or what one view shows, as csv or jsonl. It costs nothing.

get_export: checks an export and gives the download link when it is complete. The link works for an hour at most.

estimate_trigger_cost: prices a run without running it or spending anything.

trigger_rows: starts enrichment on rows and returns a runId straight away. It spends from your wallet. Without rowIds it runs every row in the table.

get_run: reports on one run.

get_status: reports progress for the whole table, including other runs.

list_run_items: shows each row's state in one run.

list_runs: lists your workspace's runs, newest first.

stop_run: stops all ongoing processing for a table. It does not refund work that was already charged.

cancel_run: cancels one run by its runId. It removes that run's queued work.

get_waterfall_trace: explains why a waterfall cell is empty or holds something unexpected, including every provider the waterfall tried, in order. It only reads. It never re-runs anything.

list_functions: lists the functions that are published, active and open to agents.

run_function: runs a function on one input or a list of up to 1,000. It spends from your wallet.

get_function_run: checks a function run. Poll it until status is no longer running.

list_capabilities: lists the enrichment actions you can run without a table. The list is the same for every workspace.

run_capability: runs one enrichment action on one input or a list of up to 1,000, with no table. It spends from your wallet unless it runs on your own connection.

create_function: creates a function as a draft.

get_function: shows one function in any state, drafts included.

open_function_sandbox: opens a sandbox, which is a copy of the function's steps that you edit. What callers run stays the same until you publish.

add_function_input, update_function_input and remove_function_input: change the function's inputs. They need an open sandbox.

add_function_step, update_function_step and delete_function_step: change the function's steps in the open sandbox. Only some kinds of step can be added this way. Anything else is refused by name and nothing is stored.

set_function_outputs: sets the whole list of columns the function sends back.

publish_function: makes the draft live. A live publish needs an admin key. Call it with dryRun: true first, which an editor key may do.

set_function_agent_access: opens a function to agents, or closes it. It needs an admin key.

delete_function: moves a function to Trash for 30 days.

list_search_filters: lists the filters for people or companies and the values each one accepts. Call it before you search.

search_people: finds people. It returns up to 50 records per call and is free to call.

search_companies: finds companies. It returns up to 50 records per call and is free to call.

create_webhook: registers an https URL. The only event today is run.settled, and the body never holds cell values.

list_webhooks: lists your outbound endpoints. It never shows the whole URL or the signing secret.

test_webhook: sends one signed ping now and reports what your receiver answered.

list_webhook_deliveries: shows an endpoint's delivery log. It is kept for 30 days.

delete_webhook: removes an endpoint and its delivery log.

get_table_webhook: tells you whether a table has an inbound address, by its last four characters. It never returns the URL.

rotate_table_webhook: makes a new inbound address for a table. The previous address stops working immediately. The new URL is shown only in this answer, so keep it.

list_schedules: lists a table's schedules. Times are in UTC.

create_schedule: runs chosen AI, API or task columns daily, weekly, monthly or at an interval of at least one hour. It spends from your wallet every time it fires, with nobody watching, until you pause or delete it. Use paused: true to save one without starting it.

update_schedule: changes a schedule. Setting paused: false starts it again, and it spends from your wallet when it fires.

delete_schedule: deletes a schedule. A run it already started is not stopped.

create_share_link: makes the link. It can expire anywhere from 60 seconds to a year after you make it, or work until you revoke it.

list_share_links: lists the live links, newest first.

revoke_share_link: stops a link at once. A revoked link cannot be turned back on.

get_wallet: shows how much the next run can spend, your plan and any spend limit on the workspace. Amounts are in US dollars.

get_usage: shows what the workspace actually spent, by table, by column or by day.

list_connections: lists the accounts your workspace can run task columns on.

test_connection: asks the provider whether a saved key is still accepted. It never returns the key.

remove_connection: removes a connection that no table uses. It refuses OAuth connections such as HubSpot and Slack. You remove those in the app.

list_provider_defaults: shows which account each provider runs on when a column names none.

set_provider_default: chooses that account for a provider. Choosing Loopster's account makes the columns you create afterwards run on it and spend from your wallet. Existing columns keep the account they were saved with.

clear_provider_default: removes that choice.

list_trash: lists what Trash holds: tables deleted on their own, workbooks and functions. Each one shows the date it is deleted for good.

restore_from_trash: brings an item back by its id. A workbook comes back with its tables. Nothing runs and nothing is billed.

send_feedback: sends a message to the Loopster team about a problem with the API. It does not open a support ticket, and nothing answers it in the conversation. Do not put secrets or personal data in it. To talk to a person, click Need Help? at the top right of any page in the app.

maxCostUsd is a ceiling, not a forecast. It comes with a breakdown for each column.

The wallet block shows spendableUsd, how many rows your wallet covers (rowsCovered) and whether the run would start at all (wouldStart).

Columns whose price cannot be known before the run are left out of maxCostUsd. unpricedColumns says how many.

Run conditions are not evaluated, so fewer rows than quoted may run, never more.

Rows whose cells already hold a finished answer are left out, unless you ask for force.

trigger_rows runs enrichment on rows. Without rowIds it runs every row in the table.

run_function bills every step inside the function as it would bill in a table, for every input.

run_capability bills each input as the same column would bill in a table, unless it runs on your own connection.

create_schedule spends every time the schedule fires, with nobody watching, until you pause or delete it. update_schedule with paused: false starts a schedule again, and it spends when it fires.

create_table spends when you pick a starter type other than blank, webhook or search.

insert_rows spends when the table's Auto-run is on, because inserting then starts enrichment. Rows that another system sends to a table's webhook address are handled in the same way.

update_row and update_cells: their descriptions say that on a table with auto-rerun columns, writing a cell re-queues the columns that read it, which spends from your wallet.

set_provider_default with Loopster's account makes the columns you create afterwards run on Loopster's account and spend from your wallet.

search_people and search_companies.

estimate_trigger_cost, which prices a run without running it.

export_table.

import_file. Importing starts no enrichment.

rename_table.

restore_from_trash. Nothing runs and nothing is billed.

create_view, update_view, create_share_link and list_provider_defaults.

set_table_auto_dedupe and set_table_create_columns_from_keys. The switch itself runs and spends nothing.

create_table for a blank, webhook or search table.

insert_rows on a table with Auto-run off. It only writes rows, and nothing runs or spends until you call trigger_rows.

If a run is bigger than your wallet, rows run in order while the wallet covers them. The rest are refused, not run. Their cells show a not-enough-balance error, and nothing is charged for them. The wallet is never overdrawn.

If the wallet cannot cover even the first row, the call is refused before anything is queued.

"List my Loopster tables and tell me which ones have Auto-run on."

"Show me the columns in {table name} and how each one is set up."

"Search Loopster Data for people with the job title Head of Sales at companies with 51-200 employees. Show me 25. Do not add anything to a table yet."

"Create a blank table with Auto-run off and add those people to it."

"What would it cost to run every column on all rows in {table name}? Do not run anything."

"Run {column name} on the first 5 rows only, then show me the results and any errors."

"The {waterfall column name} cell on the first row is empty. Show me which providers were tried."

"How much can my next run spend, and what did my workspace spend by day?"

"Export {table name} as a CSV and give me the download link."

"Make a public, read-only link to {table name} that stops working after 7 days."

"Set up a schedule on {table name} that re-runs {column name} every Monday at 09:00 UTC. Save it paused."

Create or revoke an API key. There is no tool for it. Open Settings, then Loopster API.

Add a connection. Open Settings, then Connections, then click Create.

Remove an OAuth connection such as HubSpot or Slack. remove_connection refuses these. You remove them in the app.

Create a context. Open Settings, then AI context.

Rotate the signing secret of an outbound webhook endpoint. This is not offered as a tool, so a new secret never passes through an AI model. Use the API call POST /v1/webhook-endpoints/{endpointId}/rotate-secret instead.

create_function, open_function_sandbox, set_function_outputs

add_function_input, update_function_input, remove_function_input

add_function_step, update_function_step, delete_function_step

create_share_link, list_share_links, revoke_share_link

set_provider_default, clear_provider_default, remove_connection

get_table_webhook, rotate_table_webhook

set_function_agent_access

publish_function, when it makes a function live. A dry run works with an Editor key.

delete_function, for a published function. A draft needs only an Editor key.

For insufficient_balance, add funds or ask for less. Click the wallet button at the top right, then Manage wallet. Settings opens on Plan & billing, on the Wallet tab, which has a Top up button. See Your wallet and keeping spend under control.

For workspace_limit, open Settings, then Workspaces, click the workspace and change its Spend limit with Set spend limit. Or ask for less. See Workspaces: create one and set a spend limit.

For daily_cogs_limit, topping up does not lift it, and asking for less does not help. It resets at midnight UTC, and Loopster support can raise it. Click Need Help? at the top right. See Contact support.

started: the run was handed over and at least one column runs.

none: nothing was meant to run. Auto-run may be off, or the table may have no column that an automatic run starts. An AI column needs its prompt, for example.

not_started: Loopster could not hand the run to its workers.

unknown: the run was sent but not confirmed, so the rows may or may not be running.

On not_started, call trigger_rows with the ids in rowsToTrigger. Do not insert again, because the rows are already in the table.

On unknown, call get_status before you trigger, so the rows are not run twice.

On none, check Auto-run with get_table. With Auto-run off, nothing runs until you call trigger_rows. Also check the column. See Why didn't my column run?.

deduplicated rows already had a run in progress. Nothing extra is charged for them.

rejected rows were refused because the queue is full. They did not start and will not start on their own.

unavailable rows could not be confirmed, so they may or may not have started.

rate_limited: any of the rate-limit windows.

search_concurrency: too many searches are running at once.

search_busy: Loopster Data itself is busy.

provider_busy: a create_table starter found a provider on Loopster's account at its limit, or paused after a burst of its errors. Nothing was charged.

Over MCP, the table setting that the app calls Auto-run is called autoplay. get_table shows it.

These tools take a column's name as shown in the table, or its id: update_column, delete_column, update_cells, update_row, trigger_rows and get_waterfall_trace.

Six read tools take a response_format: list_tables, get_table, list_columns, list_rows, get_row and get_status. detailed is the default and returns the full answer. concise keeps answers short. Use detailed before update_column, because changing a column needs its current setup.

To retry safely, give the call an idempotencyKey. These tools take one: create_table, insert_rows, update_row, update_cells, trigger_rows, run_capability and run_function. If you repeat a call with the same key within 24 hours, you get the first answer back, marked replayed: true, and nothing runs or bills twice. Use a new key for a new action.

Adding to a cell with update_cells in append mode is not safe to retry. A retried append adds the text twice, so send an idempotencyKey whenever you may retry.

delete_table and delete_function go to Trash for 30 days. delete_rows, delete_column and delete_view do not go to Trash. Ask your AI tool to check with you before it deletes them.

A run is kept for 30 days, and a workspace keeps its 500 most recent runs. After that, the runId is not found.

The docs list every tool with its arguments. See MCP Tools.

Connect Loopster to Claude Code and other AI tools (MCP)

Get started with the Loopster API

Use, rotate and revoke a Loopster API key

API limits, errors and retries

Create tables and add rows through the API

Run columns through the API and read the results

Troubleshoot the Loopster API and MCP