Everything machines need to work with Vipra Software and VipraGo: an OpenAPI-specified REST API, an official MCP server so AI agents can act through governed approval chains, signed webhooks, and machine-readable site maps. Read-first by design, audited by default.
Employees, attendance summaries, leave balances and requests, payroll dry-run previews, skill search. Read-first; writes flow through real approval chains; every call audited.
→ API reference (interactive) → openapi.yamlConnect Claude or any MCP-capable agent to a VipraGo org. Eight tools, zero stored state, policy-governed writes — the agent can ask, never approve its own asks.
→ GitHub: viprago-mcp-server → Quick start & tool listSubscribe to attendance.anomaly, leave.decided, payroll.completed, employee.onboarded/offboarded. Signed deliveries, 24h retries with backoff.
→ Webhook signing guideCurated entry points for LLMs and agents on both domains — plus full-text corpora for RAG systems.
→ viprasoftware.com/llms.txt → viprago.com/llms.txt → llms-full.txt (this site)Open-source home: the MCP server today; client SDKs (Node, Python) and data-engineering tooling next. Stars and issues welcome.
→ Organization profileSubscribe to engineering articles and product changes.
→ Articles RSS (this site) → VipraGo blog RSS → VipraGo changelogAI agents are becoming the new integration consumers — and giving an agent unguarded write access to payroll would be malpractice. So every Vipra developer surface follows the same contract: reads are broad, writes are requests. An agent (or your script) can submit a leave request, but VipraGo's policy engine and approval chains decide it; it can preview payroll to the rupee, but execution stays behind human confirmation. Every call — human, script, or agent — lands in the same immutable, HMAC-sealed audit log. That is what makes it safe to say yes to agents.
API keys are created in VipraGo's Control Center (admin role). Questions, scopes, or partnership ideas — talk to an engineer.
Contact Us → Open API Reference