Affinity
Search your CRM, prep for meetings and update deal records from your AI assistant.
Search records, trace audit trails and post journal entries to a DualEntry ledger.
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 DualEntry MCP server connects a DualEntry general ledger to Claude, ChatGPT and any MCP-compatible agent. Eleven of its 15 tools read — searching records, tracing audit history, matching bank lines — and four write, creating journal entries, bills, entities, revenue recognition contracts and fixed assets. Writes commit immediately. Sign-in is OAuth 2.1 through DualEntry's own login.
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.
save_record, save_entity, create_contract or create_fixed_asset. Its own troubleshooting entry for "Claude posted before you were ready" is to reverse the entry in DualEntry and add "do not post yet" to future prompts.preview_contract_schedule and preview_fixed_asset_schedule exist; there is no equivalent for save_record or save_entity.Missing permission: and writes nothing.save_record returns is the display number shown on the entry, and that passing the returned record_number to get_record_by_number can resolve a different record entirely./public/v1/, /public/v2/) authenticated with X-API-KEY, documenting a 50-request aggregate burst refilling at 5/second and a 10-request per-endpoint burst refilling at 1/second, with X-RateLimit-Limit headers and a Retry-After on a 429. That page does not cover the MCP server, and our unauthenticated probe returned no rate-limit headers, so no MCP quota is documented either way.inputSchema and no readOnlyHint for any tool. Names come from Anthropic's directory listing; behaviour comes from DualEntry's published catalogue.Straight to the ledger. DualEntry's own documentation warns that save_record posts a real record immediately, with no draft stage, no preview and no confirmation prompt between the model deciding to call it and the entry hitting your general ledger. The same warning covers save_entity, create_contract and create_fixed_asset (DualEntry, 2026).
Yes, but not through the connector. DualEntry documents Create Reverse in the dashboard, which posts a linked offsetting entry dated on or after the original. No MCP tool reverses, edits, voids or deletes anything. DualEntry also states journal entries cannot be deleted at all, so archiving is the only retirement path (2026).
They create. DualEntry documents save_record as creating an accounting record and save_entity as creating an entity, and the permission table gates both on Create rather than Edit permission on the record or entity type. Neither tool is documented as updating an existing record, so treat the verb save as create-only (DualEntry, 2026).
Yes. DualEntry documents preview_contract_schedule and preview_fixed_asset_schedule as computing the schedule a payload would produce without creating anything, and states that the preview tool takes the same payload as create_contract. That makes it a genuine dry run you can run against the exact object you intend to commit (DualEntry, 2026).
No. DualEntry states that MCP permissions match your DualEntry role, so if you cannot see something in the dashboard you cannot see it through MCP either. Checks run per record type and per entity type, and a tool you lack permission for returns a message beginning Missing permission: rather than returning the data (2026).
Yes. DualEntry documents that connectors require a Claude Pro, Max, Team or Enterprise plan, and that free plans cannot add them from the directory or as custom connectors. On Team and Enterprise plans an Owner or Primary Owner must add DualEntry for the organization first, and every member then signs in individually (2026).
Yes. DualEntry documents an append-only audit trail capturing record creation, field-level before and after values, and status changes, each stamped with the actor that caused it. Neither users nor administrators can modify or delete audit events. The connector also exposes get_record_history, which returns the audit trail for a single record (2026).
The feature is not enabled for your organization. DualEntry documents this as a subscription question rather than a permission one, and directs it to your account team. Permissions are not the only gate: revenue recognition and some record types need the corresponding subscription enabled before the tool returns anything (2026).
.md affordance; 201 words, and it is an index page that links to the four richer guides below) · retrieved 2026-09-08robots.txt carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — synthesis into AI-generated output is expressly permitted by the publisher. · retrieved 2026-09-08https://api.dualentry.com/mcp — anonymous initialize returned HTTP 401 with a WWW-Authenticate: Bearer challenge; the RFC 9728 descriptor at /.well-known/oauth-protected-resource/mcp and the authorization server metadata were both read (2026-09-08). An invented path under the same prefix returned 404, confirming the descriptor is not a pattern echo. No credentials were supplied and no tool was called. · retrieved 2026-09-08Connect DualEntry once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.