Integration

Instabase AI Hub integration

Instabase AI Hub automates document processing with AI apps. Create batches and upload the documents to process (up to 10 MB per call, larger files through a multipart session), run an automation app or a deployment on a batch or a folder, follow the run's status and read its extracted results (fields, classes, validations, confidence scores), read, write and delete files in the AI Hub filesystem, manage the organization's secrets (admins only; values are never returned) and read audit logs (Enterprise single-tenant organizations). Runs consume AI Hub usage units. Not exposed: the file-metadata route (it answers only in response headers) and the outdated batch-job poll route (get_job_status does the same). Authentication is an AI Hub API token (Settings, APIs) sent as a Bearer token, your organization ID sent as the IB-Context header, and the host of your AI Hub instance (aihub.instabase.com or your organization's custom domain).

Connect Instabase AI Hub How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Instabase AI Hub.Built from the 25 Instabase AI Hub tools below, on the connection you add once.

What you can do with Instabase AI Hub

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Batch, 4 operations

create_batchCreate batchPOSTCreates an empty batch and returns its id. Then add files with upload_batch_file (up to 10 MB) or a multipart session, and process it with run_app or run_deployment.
delete_batchDelete batchDELETEDeletes a batch and every file in it. Irreversible. Runs asynchronously: returns a job_id to poll with get_job_status.
get_batchGet batchGETReturns a batch's name, workspace, drive (mount point), workspace owner, creator, created and updated times and path suffix.
list_batchesList batchesGETLists batches (id, name, workspace, drive, owner, created and updated times), optionally filtered by workspace or creator, with limit and offset paging.

Secret, 4 operations

create_secretCreate secretPOSTStores a new secret in the organization's vault, for apps and custom functions to reference by alias. Organization admins only.
delete_secretDelete secretDELETEDeletes a secret from the organization's vault, by alias; apps that reference it lose it. Organization admins only.
list_secretsList secretsGETLists the organization's secrets (alias, description, allowed workspaces, creator and dates), never their values. Admins can list all; other members must pass a workspace they belong to.
update_secretUpdate secretPATCHChanges a secret's description, value or allowed workspaces; fields left out keep their current value. Organization admins only.

Batch file, 3 operations

delete_batch_fileDelete batch fileDELETEDeletes one file from a batch. Irreversible. The deletion runs asynchronously.
list_batch_filesList batch filesGETLists the files in a batch (name, full path, size, modified time, your read, write and delete permissions). Paginated: pass next_page_token as start_token while has_more is true.
upload_batch_fileUpload batch filePUTUploads a file (suggested 10 MB at most; larger through start_multipart_upload) to a batch under the given name, replacing a file of that name already in the batch.

File, 3 operations

delete_file_or_folderDelete file or folderDELETEDeletes a file or a folder in the AI Hub filesystem. Irreversible. Accepted (202) whether or not the path exists. Not supported on Google Drive connections.
read_fileRead fileGETDownloads a file from the AI Hub filesystem and returns it as a stored file. Pass Range to read only part of it.
write_fileWrite filePUTWrites a non-empty file to the AI Hub filesystem, creating missing parent folders and OVERWRITING any file at that path unless If-None-Match is *. Not supported on Google Drive connections.

Run, 3 operations

delete_runDelete runDELETEDeletes a run and, as chosen, its database data, input files (which also deletes the batch it processed), output files and logs. Irreversible. Returns one job id per deletion started; poll each with get_job_status.
get_run_statusGet run statusGETReturns a run's status (RUNNING, COMPLETE, FAILED, CANCELLED, STOPPED_AT_CHECKPOINT for human review, PAUSED), message, start and finish times, input and tags. Call again until it leaves RUNNING.
list_runsList runsGETLists runs, by default those started in the last 24 hours, filtered by app, deployment, user, status, workspace or time range; each with id, status, timestamps, input batch or folder and tags.

Multipart upload, 2 operations

finish_multipart_uploadFinish multipart uploadPOSTFinishes a multipart upload: action=commit with every part (part_num and part_id from upload_multipart_part, in order) saves the file to the batch; action=abort discards the session and its parts.
start_multipart_uploadStart multipart uploadPOSTStarts a multipart upload of one file larger than 10 MB into a batch and returns session_id and part_size. Then call upload_multipart_part once per part and finish_multipart_upload with every part.

App run, 1 operation

run_appRun appPOSTStarts an automation app run on a batch (batch_id) or a folder (input_dir); returns the run id at once. Poll get_run_status until COMPLETE, then call get_run_results. Uses AI Hub units.

Audit log, 1 operation

get_audit_logsGet audit logsPOSTReturns audit log entries (timestamp, user email, user IP, log type) filtered by type, user and time range; personal data. Enterprise-tier single-tenant organizations only, with audit log access granted by Instabase Support.

Deployment run, 1 operation

run_deploymentRun deploymentPOSTStarts a deployment run on a batch (batch_id), a folder (input_dir) or the deployment's upstream integration; returns the run id at once. Poll get_run_status until COMPLETE, then call get_run_results. Uses AI Hub units.

Job, 1 operation

get_job_statusGet job statusGETReturns the state of an asynchronous file job (COMPLETE, FAILED, CANCELLED, RUNNING or PENDING) and its message, such as a batch or run deletion. Call again until COMPLETE, FAILED or CANCELLED.

Multipart upload part, 1 operation

upload_multipart_partUpload multipart partPUTUploads one part of a multipart session: pass the whole file each time with first_byte and last_byte; only that slice is sent. Part n covers bytes (n-1)*part_size to n*part_size-1 (the last part can be shorter).

Run result, 1 operation

get_run_resultsGet run resultsGETReturns the extracted results of a completed run: per file and document, the class and the fields (name, value, type, error), plus packet-level fields. Paginated: when has_more is true, call again with a higher file_offset.

Use Instabase AI Hub from Claude Code, Cursor or Codex

Connect Instabase AI Hub in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Instabase AI Hub operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect Instabase AI Hub

Create your workspace
Sign up free. Nothing to install.
Connect Instabase AI Hub once
Add your Instabase AI Hub connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.