Skip to main content

ClawQL documentation

ClawQL registers core tools search and execute, audit and cache (always on), plus ingest_external_knowledge (bulk Markdown + optional URL fetch), and vault memory_ingest / memory_recall by default (set CLAWQL_ENABLE_MEMORY=0 to opt out of vault memory tools). Optional sandbox_exec (CLAWQL_ENABLE_SANDBOX=1) adds isolated code execution (Cloudflare bridge, Seatbelt, or Docker — mcp-tools § sandbox_exec). execute covers OpenAPI 3, Swagger 2, and Google Discovery via an internal OpenAPI→GraphQL projection where applicable, and optional native GraphQL / gRPC upstreams when CLAWQL_GRAPHQL_SOURCES / CLAWQL_GRPC_SOURCES are set (ADR 0002). The same tool set is available over stdio, Streamable HTTP, and (when enabled) MCP over gRPC via mcp-grpc-transport—see Deployment and Kubernetes. Full tool reference (repo).

audit and cache (ClawQL Core, no opt-out) are in-process — Core audit ring buffer vs LRU scratch — not the vault — enterprise-mcp-tools.md (#89), Session cache, cache-tool.md. Optional schedule (CLAWQL_ENABLE_SCHEDULE) adds persisted jobs and synthetic checks — Schedule synthetic checks, schedule-synthetic-checks.md (#76). Optional notify (CLAWQL_ENABLE_NOTIFY) posts to Slack — Slack notify, notify-tool.md (#77). Optional hitl_enqueue_label_studio (CLAWQL_ENABLE_HITL_LABEL_STUDIO) sends tasks to Label Studio and accepts review webhooks — HITL — Label Studio, hitl-label-studio.md (#228). Optional knowledge_search_onyx (CLAWQL_ENABLE_ONYX) queries Onyx document search — Onyx knowledge search, onyx-knowledge-tool.md (#118). Optional ouroboros_* tools (CLAWQL_ENABLE_OUROBOROS) embed the Ouroboros library evolutionary loop, with durable Postgres lineage via CLAWQL_OUROBOROS_DATABASE_URL (or split CLAWQL_OUROBOROS_DB_*) — clawql-ouroboros.md (#141, #142).

Getting started

Install clawql-mcp, run it on stdio, then point your MCP client (e.g. Cursor or Claude Desktop) at the server. Optionally set CLAWQL_SPEC_PATH, CLAWQL_SPEC_URL, or CLAWQL_DISCOVERY_URL; if you set nothing, ClawQL loads a bundled multi-provider merge (see spec configuration — default all-providers includes Google Cloud, Cloudflare, GitHub, Slack, document APIs, Onyx, and more).

Guides

Case studies

Reference