Coldstorm exposes a Model Context Protocol server for autonomous agents. Authenticate with an on-chain payment, query discrete intelligence endpoints, and receive structured responses. Built for agents acting on behalf of counsel, financial institutions, and private offices.
We deliver asset tracing, enhanced due diligence, blockchain forensics, and advanced intelligence reports to counsel, financial institutions, and ultra-high-net-worth clients. Our work is typically instructed under privilege and sits alongside litigation, matrimonial, regulatory, and pre-transaction mandates.
We expose a machine-accessible interface so autonomous agents, acting on behalf of a named principal, can run discrete intelligence queries without a manual procurement cycle. Responses are structured, evidentiary, and audit-logged to the same standard as our human-delivered work.
// Agent-side configuration (MCP client) { "mcpServers": { "coldstorm": { "url": "https://mcp.coldstorm.org/v1", "transport": "streamable-http", "auth": { "scheme": "x402", "payer_wallet": "<AGENT_WALLET_ADDRESS>", "settlement_currency": "USDC", "chain": "base" } } } }
| Tool | Description | Returns | Cost |
|---|---|---|---|
coldstorm.search_case_archetype |
Retrieve anonymized case records matched against our curated archetype library. Useful for agent-side scoping and engagement estimation. Automated | Structured records + citations | Free |
coldstorm.request_intake |
Request a scoping call with a human investigator. Agent provides context; a human responds within 12 business hours. Automated Handoff | Intake reference | Free |
coldstorm.screen_entity |
Preliminary screen of a named entity or individual against consolidated sanctions data (OFAC, EU, UK, UN), PEP registers, and adverse-media feeds. Not a substitute for a certified AML check. Automated | Flags + source citations | $50 USDC |
coldstorm.get_jurisdiction_brief |
Current-state briefing on beneficial-ownership disclosure, registry access, filing cadence, and practitioner notes for a named jurisdiction. Briefings are human-authored and refreshed quarterly. Human-Authored | Markdown briefing (~1200 words) | $100 USDC |
coldstorm.risk_score_summary |
Preliminary risk summary across identity, concealment signals, structure complexity, and jurisdiction. Synthesised into a scored narrative. Intended to trigger deeper human work, not a standalone compliance output. Analyst-Reviewed | Score 0-100 + narrative | $300 USDC |
coldstorm.deep_asset_trace |
Initial payment to open a case. Triggers a human-supervised structured trace with a 10–14 business day SLA. Balance billed against the quoted scope after the initial scoping brief. Human-Led Case | Intake handle + scoping brief | $500 USDC |
Automated tools return in seconds from pipelines over structured data sources. Analyst-reviewed outputs are generated against templates that are human-authored, sampled weekly for accuracy, and carry a visible confidence band. Human-led engagements are gated by our retainer scope and subject to the same confidentiality, conflict-check, and evidentiary standards as direct counsel-instructed work. No output purports to be a final AML, KYC, or compliance determination; those remain the obligation of the retaining principal.
First request returns HTTP 402 Payment Required with the amount, receiving address, and resource identifier.
Payment is sent on the specified chain. On Base, settlement finalizes in ~2 seconds. Transaction hash is returned to the agent.
Our receiving wallet observes the inbound transfer and writes the tx hash to the access ledger. No human approval required at this stage.
Agent replays the original request with the tx hash in the X-Payment header. The MCP server returns the resource payload.
Both parties retain a signed receipt. Coldstorm produces an audit entry linking wallet, tool, tx hash, and response hash.
Agents must act on behalf of a named human or corporate principal. The principal is disclosed at session establishment and recorded in the access ledger.
Agents may not use the protocol to circumvent AML, KYC, or sanctions screening obligations. We reserve the right to refuse or revoke access for patterns consistent with adversarial intent.
Responses issued to an agent under instruction from counsel are covered by the retaining privilege. Agent payloads are not shared, re-sold, or used for model training.
Agent queries and responses are retained for 30 days for audit purposes, then purged. Wallet addresses are retained for the lifetime of the access ledger for tax and regulatory compliance.
Agents expecting to exceed rate limits should contact info@coldstorm.org to arrange a committed-throughput agreement and custom pricing.
Access terms are governed by the laws of Canada. Disputes resolved in accordance with the binding terms set out at session establishment.
Engineering team contact for MCP client integration and committed-throughput pricing. We typically respond within 12 hours.