knowl
Service-first LLM wiki knowledge service for MCP, HTTP, and Go/Fx embedding
Install / Use
claude mcp add baldaworks -- npx -y github:baldaworks/knowlIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
AI & Machine LearningSupported Platforms
Skill content
View source on GitHubKnowl
Durable project knowledge for agents.
Knowl is a self-hosted knowledge sidecar for agentic applications. It turns durable sources into an inspectable Markdown knowledge base and returns bounded, provenance-backed evidence.
Quickstart · Connect an agent · Documentation · GitHub
What Users Get
| Capability | Result | | --- | --- | | Durable ingestion | Accepted source revisions survive restarts and remain traceable | | Inspectable knowledge | Semantic pages live in a portable, Git-reviewable Markdown workspace | | Grounded retrieval | Every response is bounded evidence with source provenance | | Agent integration | The same contract is available over MCP, HTTP, or embedded Go |
Why Knowl
- Built for durable knowledge: keep accepted project knowledge beyond one chat or agent run.
- Human-inspectable: review the generated Markdown instead of hiding knowledge in an opaque memory store.
- Clear ownership: the host selects durable inputs and writes the final answer; Knowl maintains knowledge and retrieves evidence.
- Self-hosted: run it beside an agent with your own storage and maintainer provider.
- Recoverable: durable operations and local state resume safely after process restarts.
Knowl is not session memory, a workflow orchestrator, or the primary final-answer generator. Knowl does not answer the user itself.
Minimal Sidecar Quickstart
The quickstart runs the published
v0.5.0 image with a
checked-in example source. It requires Git, Docker Compose, curl, an OpenAI
API key, and a model available to that key.
git clone https://github.com/baldaworks/knowl.git
cd knowl
export OPENAI_API_KEY='your-api-key'
export OPENAI_MODEL='a-model-available-to-your-account'
export KNOWL_OPERATOR_TOKEN='replace-with-a-local-secret'
docker compose -f deploy/sidecar/quickstart.compose.yaml up -d
curl -sS http://127.0.0.1:8080/readyz
The configured source synchronizes on startup. Check maintenance status:
docker compose -f deploy/sidecar/quickstart.compose.yaml \
exec knowl knowl --config-dir /etc source status engineering
Then retrieve grounded evidence:
curl -sS --get \
-H "Authorization: Bearer ${KNOWL_OPERATOR_TOKEN}" \
--data-urlencode 'query=Engineering shared page' \
http://127.0.0.1:8080/v1/retrieve
A successful response contains a non-empty evidence array with source
provenance. /readyz confirms that the service and storage are ready; use
source status to confirm that model-backed maintenance completed.
Stop the example without deleting its persistent volume:
docker compose -f deploy/sidecar/quickstart.compose.yaml down
The quickstart uses the hosted openai maintainer provider. Other
configurations may use opencode_acp, which requires opencode acp on PATH
and an authenticated OpenCode session. See
configuration and operations for provider and source
settings.
Connect an Agent
MCP Streamable HTTP is the primary agent-facing interface. Adapt these fields to your MCP client:
{
"transport": "streamable_http",
"url": "http://127.0.0.1:8080/mcp",
"headers": {
"Authorization": "Bearer <operator-token>"
}
}
Knowl exposes three MCP tools:
knowl_retrievereturns bounded evidence;knowl_ingestsubmits durable content or a source reference;knowl_operationreports durable operation status.
The equivalent HTTP endpoints are GET /v1/retrieve, POST /v1/ingest, and
GET /v1/operations/{operation_id}. See the
OpenAPI contract for request and response schemas.
How It Works
sources → Knowl → grounded evidence → host agent → final answer
| Component | Owns | | --- | --- | | Host agent or application | Selecting durable inputs, orchestrating tools, and generating the final answer | | Knowl | Preserving source revisions, maintaining Markdown knowledge, and retrieving evidence | | Maintainer provider | Proposing semantic updates through Knowl's validated write path |
Knowl stores immutable accepted source revisions under raw/ and semantic
knowledge under wiki/. Source documents are never copied into wiki/.
Initial bootstrap and automatic on_start synchronization are both optional.
The workspace remains inspectable and portable, with SQLite at
.knowl/knowl.sqlite providing the default local operational store:
workspace/
├── schema.md
├── raw/
├── wiki/
│ ├── index.md
│ └── ... semantic pages
└── .knowl/
└── knowl.sqlite
See the source-to-wiki showcase for a complete checked-in example.
Run Knowl Your Way
Sidecar
Use the published container for a standalone MCP/HTTP service. Persist
/var/lib/knowl, keep source mounts read-only, and pin an immutable image
digest in production. The sidecar guide covers deployment,
storage, authentication, and health checks.
Go library
pkg/knowlprovides plain-Go runtime composition;pkg/knowlfxadds Fx lifecycle integration;pkg/knowl/mcpprovides the MCP adapter;pkg/knowl/typescontains transport-neutral domain types.
Embedding changes composition, not the business contract. See the product design for architecture and ownership boundaries.
Documentation By Goal
| Goal | Start here | | --- | --- | | Deploy the sidecar | Sidecar deployment | | Configure providers, sources, and recovery | Operations guide | | Understand workspace and provenance semantics | Workspace guide | | Export the wiki for publishing or agents | OKF and llms.txt export | | Understand the architecture | Product design | | Integrate over HTTP | OpenAPI contract | | See source documents become a wiki | Source-to-wiki showcase | | Review the latest release | v0.5.0 release notes |
Contributing
Contributions are welcome. Start with CONTRIBUTING.md for local setup, verification commands, and repository conventions.
License
Knowl is released under the MIT License.
Related Skills
claude-mem
93.7kPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Understand-Anything
82.1kGraphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
Agent-Reach
79.5kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
72.2k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
