REST API
Search passages, get written answers with citations, or read whole articles and skills.
# AVIDO=https://app.avido.devcurl $AVIDO/api/v1/search \ -H "Authorization: Bearer $KEY" \ --json '{"query": "refund window"}'Give your apps and agents the same knowledge your team approved, over a REST API, an MCP server, or a folder on disk.
curl https://app.avido.dev/api/v1/answer \ -H "Authorization: Bearer $KEY" \ --json '{"question": "How long do customers have to ask for a refund?"}' # 200 OK{ "answered": true, "answer": "Customers can ask for a full refund within 30 days [1].", "citations": [ { "n": 1, "title": "Refund policy", "approved": true } ]}Avido does the hard part: keeping knowledge right. You build the experience on top.
Answer customers from approved knowledge only, with citations anyone can check.
POST /answerPut approved passages into your own prompts, with the model you already use.
POST /searchCopy passages and their embeddings into the store you run, and keep it in sync.
GET /chunks · webhooksConnect Claude or any MCP client. Each person sees only what is shared with them.
MCP serverPull knowledge and skills into a repo on a schedule, and point Claude Code at the skills.
CLI · skillsShow approved articles on your own site, and rebuild a page when its article is published.
GET /articles · article.published// Ground your own model in approved knowledgeconst res = await fetch(`${AVIDO}/api/v1/search`, { method: "POST", headers: { Authorization: `Bearer ${KEY}`, "content-type": "application/json", }, body: JSON.stringify({ query: question, topK: 5 }),});const { data } = await res.json(); const context = data .map((p) => `# ${p.title}\n${p.content}`) .join("\n\n"); // Send `context` and `question` to your model.// Every passage is approved unless you ask otherwise.Search passages, get written answers with citations, or read whole articles and skills.
# AVIDO=https://app.avido.devcurl $AVIDO/api/v1/search \ -H "Authorization: Bearer $KEY" \ --json '{"query": "refund window"}'Plug any MCP client into the same knowledge. Agents use a key; people sign in as themselves.
claude mcp add --transport http \ avido $AVIDO/mcp # Tools: search_knowledge,# get_article, list_articles,# list_skills, get_skill,# report_problemPull approved articles and skills into a folder as Markdown. Later pulls only change what changed.
npx avido-knowledge pull ./knowledge # knowledge/# articles/<collection>/<slug>.md# skills/<name>/SKILL.mdKeys read approved content only. Opt in to content that is not reviewed yet, per request.
Each agent gets its own key and sees only what is shared with it. Rotate or disable it any time.
Every article, passage, and answer says whether it is approved or not reviewed yet.
Keep your own index, cache, or vector store in step. Deliveries are signed and retried for hours, and the same events can be read back from an ordered feed.
// Verify a deliveryconst expected = hmacSha256Hex( secret, `${headers["x-knowledge-timestamp"]}.${rawBody}`,);if (expected !== headers["x-knowledge-signature"]) reject();Nothing an app or agent sends goes live on its own. It goes to Review, where a person decides.
An agent can say an article is wrong, outdated, unclear, or missing. It lands in Review, with a fix drafted.
POST /problemsTell Avido what your AI must be able to answer. Anything not covered shows up as a gap.
POST /questionsWord, PDF, Markdown, and text in. Each section becomes an article draft for review.
POST /uploadsSend a SKILL.md with its files. A new version waits for a person to approve it.
POST /skillsBearer keys, cursor paging, and plain JSON. The full reference, with examples in curl, JavaScript, and Python, is in the app.
Join the waitlist to get a workspace and your first key, or book a demo and we will walk through your use case.