ClawQL — the Agentic Gateway for Auditable Production AI

ClawQL is the Foundational Platform where agents search APIs, execute operations, recall vault knowledge, ingest documents, and audit what happened — from multi-tenant Regional Hubs to Dedicated Virtual Gateways and Edge Gateways on every laptop.

Why ClawQL exists

Built from production agent work

ClawQL started as the MCP layer behind real deployments: multi-provider DevOps workflows, document pipelines for regulated industries, and cross-session memory that survives when you switch between Cursor, OpenClaw, and Kubernetes. The open-source core ships what we run in production.

ClawQL and See The Greens

See The Greens is a mortgage-first loan origination system powered by ClawQL — the same IDP pipeline, vault memory, and HITL patterns documented on this site. ClawQL is the horizontal platform; See The Greens is one vertical product built on it. Lenders can self-host ClawQL or build their own LOS on the same stack.

Current state

The open-source MCP server is production-ready and self-hostable today — npx, Helm, published case studies, and a v7 operator for Kubernetes. Managed gateway hosting is early access (waitlist, then paid tiers). IDP hosted tenants are onboarding with founder-led setup. We are pre-revenue on managed hosting; the core product is shipping and in use.

First buyers

Document-heavy teams of 20–200 people in legal, M&A diligence, healthcare operations, and lending — teams that already spend $500–2k/mo on SaaS and need agent memory plus document intelligence. Gateway-only developers evaluating MCP infrastructure start on Developer or Teams; IDP activates when they need classify/extract and VDR.

Early access, founder-led

Managed hosting onboarding is founder-led — not self-serve checkout. Hosted slots are limited while we validate pipeline reliability under real workloads. Architecture earns trust over time; we do not claim compliance certifications or decade-long references we do not have yet.

What we publish

Case studies include tool traces, failure modes, fixes, and token measurements. The GitHub release changelog tracks what ships each version; docs at docs.clawql.com stay aligned with the codebase.

What we build

Three problems — one platform, activated in layers

Current agent systems lose context, burn tokens on API specs, and treat documents as afterthoughts. Plugin bundles let you solve gateway and memory first; add IDP and sovereign inference only when your workload needs them.

Persistence gap

Context windows are ephemeral. memory_ingest and memory_recall give agents institutional memory that survives restarts, product switches, and new chat threads.

Token efficiency

GitHub's bundled spec alone is ~2.28M planning tokens. search returns ranked operation IDs in thousands of tokens — measured in our published GitHub provider case study.

Document intelligence

Eight bundled IDP vendors compose intake through secure sharing. Agents orchestrate real document pipelines — classify, extract, archive, search — built around purpose-built extraction tools rather than one-off OCR prompts.

Tool surface

ClawQL Core

Always on — no opt-out

search

Rank thousands of API operations by natural-language intent. Specs stay server-side; agents get operation IDs and parameter hints. Provider catalog is empty by default — opt in with CLAWQL_PROVIDER / providers.pack / CLAWQL_TIER.

execute

Run one operation with validated args over REST, GraphQL, or gRPC. Lean responses keep tool results out of your context budget.

audit

In-process ring buffer of structured events — append, list, clear. Pair with memory_ingest when you need a durable operator trail.

cache

Ephemeral LRU scratch space for the active session. Use memory_* when data must survive restarts.

skills_list

Discover shipped Agent Skills (SKILL.md packs) registered on the server — including handoff and workflow recipes.

skills_get

Fetch a skill body by name so agents can follow vetted SOPs instead of improvising every workflow.

Memory & documents

Default on — opt out with CLAWQL_ENABLE_MEMORY=0 (documents follow the same default-on pattern)

memory_ingest

Write Obsidian Markdown with wikilinks and structured insights. Decisions, runbooks, and postmortems persist across chat sessions.

memory_recall

Keyword search, wikilink graph hops, optional vector KNN. Agents pull institutional knowledge rather than re-asking the same questions.

ingest_external_knowledge

Bulk Markdown or HTTPS URLs into the vault — ideal for importing docs, threads, and vendor analyses.

Automation & ops

Opt in via CLAWQL_ENABLE_* (with CLAWQL_INSTANCE_SPEC / CLAWQL_TIER composition in 8.0)

schedule

Persisted synthetic HTTP checks with assertion-based pass/fail and run history — uptime without a separate monitoring stack.

notify

Slack chat.postMessage wrapper for workflow milestones. Pipe IDP completion summaries straight to your team channel.

sandbox_exec

Isolated code snippets via Seatbelt, Docker, or Kata — validate transforms without giving agents a raw shell.

workflow / argocd

Submit Argo Workflows and observe Argo CD Applications from the same MCP surface your agents already use.

knowledge_search_onyx

Hybrid enterprise search over your Onyx index. Opt in with CLAWQL_ENABLE_ONYX (plus instance/tier composition).

IDP document pipeline

Opt in — pdf-inspector route, eight vendors, grounded extract

inspect_pdf

Firecrawl pdf-inspector — classify TextBased vs scanned, extract Markdown, and decide whether Docling OCR is needed.

run_idp_pipeline

Automated multi-hop pipeline: Nextcloud intake → Docling/Tika → Gotenberg → Stirling → archive → Onyx → Coneshare VDR.

classify_document

Route documents by type before extraction — W-2s, invoices, and contracts hit the right downstream handlers.

extract_document

Schema-grounded LangExtract fields with char_interval provenance — grounded extraction with measurable confidence.

hitl_enqueue_label_studio

Push low-confidence extractions to Label Studio for human review, then ingest decisions back into the vault.

Evidence

Case studies include tool traces, failure modes, fixes, and token measurements — the same detail we expect from production engineering write-ups.

~2.28M planning tokens → ~3.4K

GitHub without pasting 9 MB of OpenAPI

Agents search 1,099 indexed GitHub operations by intent, then execute list-commits and repos/update — the spec stays in the MCP server, not the chat.

Read case study

Control plane + Worker-safe app

Ship docs.clawql.com via MCP

Deploy Cloudflare Workers with search/execute for zones and domains — then fix OpenNext fs/runtime issues until docs.clawql.com is reliable.

Read case study

Cross-session vault recall

OpenClaw recalls prior sessions

A fresh OpenClaw agent thread calls memory_recall and returns ranked vault notes from earlier Cursor and K8s work — no copy-paste.

Read case study

Read the case studies. Run the tools.

Self-host today or request early access for managed hosting. For the default product-led motion, see the inference-first GTM playbook; for regulated / Palantir-facing positioning, see the enterprise GTM playbook. Full Apache 2.0 stack, no license fee.

Request access All case studies