Maha Context Review · Free connector documentation
Choose a smaller context pack and check whether the exact evidence you declared survived. The Claude connector is free, requires no Maha account or payment, and works on authorized sanitized or synthetic text you explicitly supply.
Find the free Claude connector
Start with a safe example
- Connect Maha Context Review in Claude and enable it for your conversation.
- Supply a task and documents with stable source IDs and nonempty text. Remove credentials, government identifiers, payment details, protected health information and confidential personal or company records first.
- Choose an estimated token budget from 64 to 16,000. Requests support up to eight sources and 64 KiB total.
- Declare the source ID and exact, case-sensitive excerpt that must survive. An empty evidence list means completeness was not tested.
Example 1: an excerpt survives
Use Maha Context Review on this synthetic text. Task: What service rebate applies? Budget: 256 tokens. Source ID: amendment. Source text: The service rebate is 25%. Required evidence: source amendment, exact excerpt 25%. The supplied text is synthetic and sanitized. Report the returned retention status and limitations.
Example 2: the right words in the wrong source
Use the same synthetic task and budget. Source amendment says: The service rebate is 10%. Source other says: The service rebate is 25%. Require excerpt 25% specifically from amendment. Report why the other source cannot satisfy that declaration.
Example 3: no evidence declared
Use source amendment: The service rebate is 25%. Task: What service rebate applies? Budget: 256 tokens. The text is synthetic and sanitized. Leave the required evidence list empty. Explain the returned declaration status without claiming completeness was tested.
Read the results carefully
The tool reports selected passages, estimated token changes, literal source-specific evidence checks and limitations. Inspect missing evidence, source coverage and warnings together. A smaller or empty pack is not automatically a useful result.
This is deterministic text selection and literal checking. It does not independently verify source authority, semantic completeness, factual accuracy, legal meaning, signatures or what a downstream model actually received. It does not browse the web, call another AI model, retrieve your accounts or take a payment.
Other clients and support
Remote MCP endpoint: https://www.mahastrategies.com/api/mcp/context-review. The published tool is review_context_retention. No Maha API key is required.
Free web workbench · Connector data handling · Connector usage boundaries
For help, email mayone@mahastrategies.com with a minimal synthetic example. Free connector support does not require buying a Sprint.
Optional: get the claims resolved
If you need source-linked findings and revised wording for a proposal or procurement response, Maha offers a separate $250 USD Claim Resolution Sprint. It covers up to three agreed claims against one designated English source of up to five pages, with a PDF/JSON packet, change log and one correction round.
This is an assisted service with a written scope and customer reviewer decision. It is not a paid unlock of the connector, an automatic tool result, certification, or guaranteed positive findings. No connector input is automatically copied into a paid engagement.
You choose whether to contact us. Scope and terms are agreed before any payment request.