Key facts
| Agent stack | Chat completions, function calling and embeddings are live |
| Use cases | Onboarding checks, fraud alert triage, transaction categorisation, reconciliation |
| Human review | Account-affecting actions and fraud decisions stay with people |
| Tools | Read-only lookups plus capped actions with approval queues |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped options |
| Models | 30+ models; small aliases for classification, stronger ones for investigations |
| Pricing model | Flat monthly self-serve plans with unlimited fair-use usage |
| Roadmap | Batch and fine-tuning endpoints are coming soon |
TL;DR
- Automate onboarding extraction and reconciliation matching.
- Enrich fraud alerts and rank them for an investigator.
- Cap every action and require approval for account-affecting changes.
- Keep an audit trail of prompts, sources and reviewer identity.
- Prototype on sanitised data before touching production records.
How it works, step by step
- Pick one workflow such as onboarding document extraction or alert enrichment.
- Expose read-only tools for documents, transactions and case systems.
- Use strict schemas for extracted fields with per-field source references.
- Set hard caps on any action tool and route exceptions to a human reviewer.
- Configure region, retention, scoped keys and audit logging before the pilot.
- Log everything, then measure analyst time saved and exception quality.
Try it yourself
Open the AI agent cost calculator →
Fintech workflows suited to agents
Fintech is document- and alert-heavy, which fits prepared, reviewable work:
- Onboarding: extract entity, ownership and document fields for a reviewer to verify.
- Fraud triage: enrich alerts with device, history and network context, and rank them for an investigator.
- Transaction categorisation: assign categories and flags with confidence scores for downstream review.
- Reconciliation: match payments and ledger entries, and list unmatched items with reasons.
Investigators and operations staff keep authority over outcomes.
Architecture with caps and review
Design for scrutiny:
- Retriever: embeddings over policy, procedures and case history.
- Extractor: strict schema with per-field references and confidence.
- Read-only tools: account, transaction and document lookups.
- Capped actions: limited operations with an approval queue above thresholds.
- Reviewer: a person confirms fraud outcomes and account changes.
Log prompts, tool calls, sources, model version and reviewer for every case so an auditor can reconstruct events.
Data, model risk and deployment
Financial data is sensitive and often subject to specific handling obligations. Minimise what enters a prompt, prefer references over raw records, and set region and retention deliberately. Use scoped keys, RBAC, SSO and audit logs, and pin model versions so behaviour is reproducible for review. Expect model risk and outsourcing questions from partners and regulators: be ready to describe how the agent assists a documented human process. Where isolation is required, deploy in a VPC, on-prem or air-gapped environment; the compatible API keeps code, prompts and evaluations unchanged.
Pilot and evaluation
Prototype on plugsky-micro or plugsky-lite with sanitised cases, then use the 14-day full-access trial to test a stronger model on complex investigations. Track analyst minutes per case, extraction accuracy, alert ranking quality, exception rate and reconciliation match precision against the current process. Self-serve plans are flat monthly with fair-use usage, so volume growth does not create billing swings. Start with one queue, keep prompts versioned, and re-run the evaluation before extending to adjacent workflows.
Honest comparison
| Capability | Plugsky | Typical agent framework | Building in-house |
|---|---|---|---|
| Tool calling | Live function calling with read-only and capped action tools | Framework-specific adapters | You build and maintain |
| Audit | Log prompts, tool calls, sources and reviewers | Partial or add-on | You build all of it |
| Deployment | Cloud, VPC, on-prem and air-gapped | Usually cloud-only | You operate GPUs and serving |
| Pricing | Flat monthly self-serve, fair-use usage | Varies by framework and model | GPU plus operations cost |
| Model choice | 30+ models behind one API | Varies by provider | You host every model |
Frequently asked questions
Can the agent freeze or close an account?
No. Keep account-affecting actions with a human reviewer; if you allow capped actions, enforce hard limits and an approval queue.
Which endpoints do fintech agents use?
Chat completions, function calling, JSON mode and embeddings are live. Batch and fine-tuning endpoints are coming soon.
How do we protect customer data?
Minimise what enters prompts, prefer references, set region and retention, and use scoped keys and audit logging.
Is there a free plan?
Yes. The free plan includes plugsky-micro and plugsky-lite with no credit card, suitable for a sanitised prototype.
How does pricing work?
Self-serve plans are flat monthly with unlimited fair-use usage. See the live pricing page for current plans and enterprise options.
Can we keep everything in our cloud?
Yes. VPC, on-prem and air-gapped deployments are available with the same compatible API, so your application code and tests stay valid.
How should we measure it?
Track analyst minutes per case, extraction accuracy, ranking quality, exception rate and match precision against the current process.