# Syntheka

> The open, governed alternative to Palantir: enterprise AI agents grounded in a typed business ontology, where every write-side action is proposed, approved, audited, and reversible. Self-hostable in one command; SAP-native; 255 OpenAPI endpoints / 368 schemas across 18 modules.

## Pages
- [Home (English)](https://syntheka.ai/): product overview, deployment options, pricing, FAQ
- [首页（中文）](https://syntheka.ai/zh): 中文镜像
- [Full site text](https://syntheka.ai/llms-full.txt): complete text of all content pages and blog essays

- [SAP integration (English)](https://syntheka.ai/sap-integration): governed write-back, three iron laws, connector coverage
- [SAP 集成（中文）](https://syntheka.ai/zh-sap-integration): 中文镜像

- [Security: will AI leak your business data?](https://syntheka.ai/security): shadow AI numbers, four isolation boundaries, egress control
- [安全：AI 会不会泄露业务数据（中文）](https://syntheka.ai/zh-security): 中文镜像

- [Cost: where does the AI budget go?](https://syntheka.ai/cost): 72% outside the invoice, six leaks, governed execution as cost control
- [成本：大模型花费都去哪了（中文）](https://syntheka.ai/zh-cost): 中文镜像

- [Platform: what's inside](https://syntheka.ai/platform): six-stage lifecycle, engineering numbers, what you own self-hosted
- [平台：Syntheka 里面有什么（中文）](https://syntheka.ai/zh-platform): 中文镜像

- [Incidents: six public agent failures, one shared anatomy](https://syntheka.ai/incidents): public incident taxonomy mapped to four missing controls — staging, approval, grounding, boundaries
- [事故：六起公开 Agent 事故（中文）](https://syntheka.ai/zh-incidents): 中文镜像

- [Construction compliance](https://syntheka.ai/construction-compliance): contract requirements extracted, certificates compared, non-compliant payments blocked with governed approvals and audit
- [Requirements fit check](https://syntheka.ai/requirements): submit your business need, get a verdict against the capability map — three marks of a deliverable requirement; FDE-level delivery via agent + MCP instead of embedded headcount
- [需求匹配测评（中文）](https://syntheka.ai/zh-requirements): 中文镜像
- [Fit check workbench](https://syntheka.ai/fit): sign in with a one-time email link, run up to three automated fit checks against the capability map and the three marks — instant initial assessment, not a delivery commitment
- [匹配测评工作台（中文）](https://syntheka.ai/zh-fit): 中文镜像
- [建筑分包合规（中文）](https://syntheka.ai/zh-construction-compliance): 中文镜像
- [E-commerce recovery](https://syntheka.ai/ecommerce-recovery): settlement reconciliation surfaces mis-collected fees and duplicate charges — recoveries staged, approved, and tracked
- [电商结算追回（中文）](https://syntheka.ai/zh-ecommerce-recovery): 中文镜像

- [Privacy Policy](https://syntheka.ai/privacy): no cookies, no tracking, plain language
- [Terms of Use](https://syntheka.ai/terms): plain-language terms, early-access honesty

- [Blog](https://syntheka.ai/blog/): essays on governed AI agents, SAP write-back, landscape analysis

## Comparisons
- [Syntheka vs Dify](https://syntheka.ai/vs-dify): where horizontal agent builders fit and where governed write-side action begins
- [Syntheka vs n8n](https://syntheka.ai/vs-n8n): workflow automation vs governed agents acting on systems of record
- [Syntheka vs myCOI](https://syntheka.ai/vs-mycoi): COI tracking software compared — passive tracking vs payment-gated compliance
- [Syntheka vs Trustlayer](https://syntheka.ai/vs-trustlayer): subcontractor risk platforms compared — document collection vs a governed verification loop
- [Syntheka vs Chargeflow](https://syntheka.ai/vs-chargeflow): automated chargebacks vs governed settlement reconciliation and fee-mischarge recovery

## Essays
- [Enterprise agents need an approval loop, not just logs](https://syntheka.ai/blog/governed-agents-approval-loop): the mechanism, frame by frame
- [Provable trust: not just watching your AI agents](https://syntheka.ai/blog/provable-trust-runtime-policy-enforcement): staged writes, runtime policy enforcement, and a replayable evidence chain
- [Governed write-back to your system of record](https://syntheka.ai/blog/kernos-sap-governed-writeback): three iron laws for touching SAP with agents
- [Open-source Palantir alternatives, honestly assessed (2026)](https://syntheka.ai/blog/open-source-palantir-alternatives-2026): a fair map of the field and the gap none of them close
- [How to verify subcontractor insurance before releasing payment](https://syntheka.ai/blog/verify-subcontractor-insurance-before-payment): a four-step procedure
- [Subcontractor compliance checklist for payment runs](https://syntheka.ai/blog/subcontractor-compliance-checklist): nine checks before every payment
- [Why COI tracking spreadsheets fail general contractors](https://syntheka.ai/blog/coi-tracking-spreadsheet-fails): three failure modes and what replaces them
- [Subcontractor prequalification: the annual questionnaire problem](https://syntheka.ai/blog/subcontractor-prequalification-process): why the snapshot fails and where the checks belong instead
- [Additional insured vs. certificate holder](https://syntheka.ai/blog/additional-insured-vs-certificate-holder): the checked box is not the endorsement
- [Approval loop essay](https://syntheka.ai/blog/governed-agents-approval-loop)
- [Provable trust essay](https://syntheka.ai/blog/provable-trust-runtime-policy-enforcement)
- [SAP write-back essay](https://syntheka.ai/blog/kernos-sap-governed-writeback)
- [Landscape essay](https://syntheka.ai/blog/open-source-palantir-alternatives-2026)
- [COI tracking spreadsheet failure modes](https://syntheka.ai/blog/coi-tracking-spreadsheet-fails)
- [Subcontractor insurance verification before payment](https://syntheka.ai/blog/verify-subcontractor-insurance-before-payment)
- [Subcontractor compliance checklist for payment runs](https://syntheka.ai/blog/subcontractor-compliance-checklist)

- [Syntheka vs Dify comparison](https://syntheka.ai/vs-dify): honest line-by-line comparison — where each wins

- [Syntheka vs n8n comparison](https://syntheka.ai/vs-n8n): automation reach vs governed write-side authority

- [Syntheka vs myCOI comparison](https://syntheka.ai/vs-mycoi): certificate tracking vs payment-time enforcement, where each wins
- [Syntheka vs TrustLayer comparison](https://syntheka.ai/vs-trustlayer): document verification vs payment-time enforcement, where each wins

- [隐私政策（中文）](https://syntheka.ai/zh-privacy)
- [使用条款（中文）](https://syntheka.ai/zh-terms)

## Facts
- Category: enterprise AI agent platform (ontology-driven, governance closed loop)
- Deployment: self-hosted Docker Compose (today); private cloud with vendor team (today); managed multi-tenant SaaS (planned)
- Integrations: SAP via OData / IDoc / BAPI / events / CDC, MDM external-ID mapping
- Storage: PostgreSQL 17 + pgvector + Lance columnar
- Pricing: Starter $500/mo (3 agents), Professional $2,000/mo (10 agents), Enterprise custom; construction units: Starter ~20 active subcontractors, Professional ~100, Pilot $1,500 flat (first project, 60 days)
- Governance: staged actions, multi-node approval DAG with segregation of duties, bi-temporal append-only audit with causal IDs
- Construction scenario: ACORD 25 certificate parsing, requirement extraction from contracts, payment-application blocking with approval DAG and audit chain
- Security posture: bounded host allowlist for outbound connectors; opt-in egress auth gate
- Contact: hello@syntheka.ai
- Last updated: 2026-10-03