Aleph
Query live FP&A data, build reports and update dashboards from your AI assistant.
Import, adapt and expand training datasets, then launch AutoScientist fine-tuning runs from an AI agent.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Adaption MCP server connects Claude and other MCP clients to Adaption Labs' data and fine-tuning platform. Its 18 tools import datasets, run Adaptive Data processing, add synthetic or translated rows, invent datasets from a description and launch AutoScientist training runs. Several tools spend organization credits. Sign-in is OAuth, scoped to one organization you choose.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
create_autoscientist_run launches without an estimate mode, per Adaption's docs. Anthropic's directory says every credit-spending tool supports an estimate; the vendor's tool reference says otherwise for this tool.Some tools do. Adaption's documentation says adaptation, augmentation, translation, localization, dataset generation and AutoScientist training can consume credits from your organization's balance. Five of those tools accept an estimate request that returns the expected cost before anything runs. Listing, reading and checking status do not launch jobs. Adaption's FAQ adds that unused credits stay on the balance and do not expire.
Yes, it can. Adaption's documentation states that the training-run tool has no estimate mode, unlike the five dataset tools that do. Its advice is to ask for recommended hyperparameters first so you can review the base model and configuration before launch. Anthropic's directory listing says every credit-spending tool supports an estimate, which the vendor's own tool reference contradicts for this one tool.
Because each sign-in grants access to one Adaption organization only. Adaption's documentation says you pick the organization on the consent screen, and the client then sees only that organization's datasets and AutoScientist runs. To switch, sign in again from your client and choose the other organization. Access also ends when you leave an organization, so a stale connection needs a fresh sign-in.
Yes, with an API key. Adaption's documentation describes sending an Adaption API key as a bearer token for machines without a browser, such as a CI runner. A key made in settings has full access by default. A key created with an explicit scope list needs the read scope to read data and the write scope to create datasets or launch jobs.
It means the credential lacks a scope the tool needs. Adaption's troubleshooting guide says this happens when an API key was created with a scope list that leaves out the read or the write scope. Reading datasets and runs needs the read scope. Importing, combining and every credit-spending job needs the write scope, which includes read. Create a key with the right scopes and reconnect.
POST initialize against https://api.prod.adaptionlabs.ai/api/v1/mcp (2026-11-18): HTTP 401 with WWW-Authenticate naming the RFC 9728 descriptor and the scopes adaption-mcp:read adaption-mcp:write. The descriptor and the RFC 8414 authorization server metadata both returned HTTP 200. · retrieved 2026-11-18llms.txt (read 2026-11-18), used to locate the pages above. · retrieved 2026-11-18docs/marketing/data/anthropic-directory-2026-11-17.json. Source for the partner tier, the 2026-10-09 listing date, the endpoint, the 18 tool names and the estimate-mode claim. · retrieved 2026-11-17Connect Adaption once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.