Bucket and object handling maps straight onto S3 — buckets created with an isPublic flag, objects addressed by bucketName and key, signed URLs carrying an expiresIn — and the credentials are the standard AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY pair with AWS_ENDPOINT_URL_S3 aiming them at Tigris, so an environment already set up for S3 needs the endpoint changed and little else. Worth knowing before those credentials go into an agent that plans its own steps: tigris_delete_bucket and tigris_delete_object sit in the same tool list as the read calls, with no separate confirmation argument on either.
An object-storage client for Tigris over the S3 API, with 10 tools covering bucket lifecycle, object read, write and delete, and signed URL generation.
- Bucket lifecycle in the same surface as the object calls: tigris_create_bucket takes a bucketName and an isPublic flag, tigris_list_buckets returns the buckets in the account, and tigris_delete_bucket removes one.
- Writing an object from memory or from disk: tigris_put_object stores inline content under a key with a stated contentType, and tigris_put_object_from_path reads the bytes from a path on the filesystem instead.
- Reading and clearing by key: tigris_get_object fetches an object from a bucketName and key, tigris_list_objects lists what a bucket holds, and tigris_delete_object removes a key.
- Links that expire: tigris_get_signed_url_object returns a signed URL for an existing key with an expiresIn, and tigris_upload_file_and_get_url collapses the upload from a path and the URL into a single call.
A Tigris account and an S3 key pair — AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY — plus the S3 endpoint to send them to, supplied as AWS_ENDPOINT_URL_S3.
One command plus a key — npx -y @tigrisdata/tigris-mcp-server init, then supply credentials