Tropic

Benchmark software pricing, search contracts and pull negotiation playbooks.

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

What the Tropic connector does

The Tropic MCP server connects a Tropic procurement workspace — software contracts, renewal dates and negotiated pricing — to Claude, ChatGPT and any MCP-compatible agent. Anthropic's directory lists 10 tools that benchmark prices, search contract documents and surface supplier negotiation playbooks. Sign-in is OAuth against your own Tropic account.

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.

Tropic tools (10)

  • get_supplier_intelligence
  • get_negotiation_playbook
  • run_price_benchmark
  • list_supplier_products
  • get_server_status
  • query_contract_documents
  • create_manual_benchmark_request
  • list_price_benchmarks
  • get_price_benchmark
  • list_contracts
  • list_expert_insights
  • list_market_insights
  • list_redundant_spend
  • get_redundant_spend

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • No application scopes exist. The RFC 9728 descriptor at https://app.tropicapp.io/.well-known/oauth-protected-resource/mcp advertises only openid, profile and email. Nothing at the OAuth grant separates reading contracts from searching their text or from queuing a benchmark request. Your Tropic account role is the only boundary.
  • Tropic publishes no tool list. Its MCP article names six capabilities and zero tools, so the 10 names come from Anthropic's directory and matched only semantically. A rename would be invisible to that diff.
  • Two documented capabilities have no matching tool. Compliance Check and Contract Insights appear in Tropic's MCP article, but Tropic's feature documentation places both inside the platform — compliance checks run through Ask AI, and Contract Insights is the Purchase Prep Assistant's output. Whether either is reachable over MCP is unresolved.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail or readOnlyHint is published here.
  • The connector cannot act on what it finds. It reads renewal dates but cannot change one, cannot opt out, cannot approve a request and cannot notify a supplier. Every action still happens in Tropic. Treat the connector as analysis, and keep the execution path human.
  • Benchmark cost is unpublished. Tropic documents no per-benchmark charge, and no tool in the surface can observe spend, credits or quota — a scan of all 10 names for billing vocabulary returns nothing. An agent cannot check an allowance before consuming one, so do not blanket auto-approve create_manual_benchmark_request.
  • A custom benchmark request reaches a person. Tropic staff build these manually and may tag the requester in the benchmark Activity Feed for more information, so an automated call creates human work at Tropic, not just a database row.
  • Contract document text is untrusted input. Supplier-drafted agreements enter the agent's context through query_contract_documents, and a supplier writes that text. Treat retrieved contract language as data, never as instructions.
  • Benchmarks must not reach the supplier. Tropic states benchmarks are internal-only and never shared with suppliers. The connector cannot enforce that, so an agent with both Tropic and an email tool is a leak path worth designing against.
  • Tropic's AI commitments do not travel. Its no-training, client-isolation and stateless guarantees describe Tropic's own AI features. Data pulled through this connector is governed by your MCP client's terms instead.
  • The vendor documentation host blocks direct browser fetches. help.tropicapp.io returned HTTP 403 to our requests while its robots.txt permits /hc/ articles, so the article was read through the same host's public Zendesk help-centre JSON API. The block is bot-filtering, not a policy signal.

Frequently asked questions

Can the Tropic connector approve a purchase or commit spend?

No. None of the 10 tools approves a request, signs a contract, issues a purchase order or sends anything to a supplier. Tropic's approval, workflow-task and purchase-order features exist in the platform and have no counterpart in this listing. The one write tool queues an internal analysis request, which spends no money and binds nobody.

Can the connector trigger or stop a contract renewal?

No, in both directions, and the second half matters more. No tool edits a contract stage, an opt-out date or a renewal reminder, so an agent cannot cause a renewal to execute. It equally cannot stop one. Reading a renewal date through Claude changes nothing in Tropic; the opt-out deadline still has to be actioned in the platform.

Can the Tropic connector modify or delete a contract record?

No. Not one of the 10 tool names carries a delete, update, edit or archive verb, and Tropic's own contract deletion is a permanent, irreversible platform action it documents as removing linked transactions from reporting. Nothing in this listing reaches that path, so the contract records finance and audit rely on stay read-only here.

Does the Tropic MCP server expose other companies' negotiated pricing?

Only in anonymized aggregate, never as a named third-party contract. Tropic states benchmarks are derived from anonymized contract data across more than $20B in spend, and that they are for internal use only and never shared with suppliers. Treat benchmark output as commercially sensitive: it is market intelligence about suppliers you negotiate against.

What OAuth permissions does the Tropic MCP server request?

Only identity claims. Tropic's RFC 9728 resource descriptor advertised scopes_supported of openid, profile and email on 2026-08-23. None of the three names a Tropic resource or a verb, so consent grants no way to allow benchmark reading while withholding contract document search. Access is governed by your Tropic account instead.

Does creating a benchmark request through the connector cost money?

Tropic publishes no per-benchmark charge, and the connector exposes no billing or quota tool to check one. A custom benchmark is fulfilled manually by a Tropic specialist, typically within one business day, so it consumes a service entitlement rather than a metered credit. Confirm your plan's benchmark allowance with Tropic before automating the call.

What do I need before connecting Tropic to Claude?

A paid Claude plan and two separate setup actions. Tropic states free Claude accounts cannot add the connector, that a Claude Workspace Owner must enable Tropic org-wide once, and that every individual user must then connect it to their own account. If Tropic is missing from your connector list, the admin step has not happened yet.

Why does Claude not call the Tropic connector even though it is connected?

Usually because the prompt does not signal Tropic explicitly. Tropic's guidance is to rephrase naming the tool, such as asking what the Tropic price benchmark is for a supplier, and to confirm the connector shows as connected under Connect your apps. Starting a fresh chat is Tropic's other documented fix for the same symptom.

Sources

Use it in an agent

Put Tropic to work.

Connect Tropic once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Call (650) 285-1019Our AI receptionist answers.