Haki MCP Server

by gethaki Your server? Claim it

answering

Haki is answering right now. Last checked 4 min ago. It exposes 5 tools. Last commit 12 Sep 2026.

Long-term memory for AI agents: bitemporal fact ledger, contradiction detection, explainability.

Uptime history 10 days of history · worst day 99%

10 days agonow

100.0%

Uptime 24h

92 of 92 checks

Tools

read from the server

221 ms

Response time

average over 24h

Stars

last commit 12 Sep 2026

Nothing serious here today

Today is the operative word: we check Haki every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.

Connect this server

Endpoint below is the one we actually reach during checks — not the one copied from a README. Last verified 4 min ago.

Claude Code Claude Desktop Codex Cursor VS Code

run in your terminal

claude mcp add haki --transport http https://api.gethaki.space/mcp/

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "haki": {
      "url": "https://api.gethaki.space/mcp/"
    }
  }
}

~/.codex/config.toml

[mcp_servers.haki]
url = "https://api.gethaki.space/mcp/"

.cursor/mcp.json

{
  "mcpServers": {
    "haki": {
      "url": "https://api.gethaki.space/mcp/"
    }
  }
}

.vscode/mcp.json

{
  "mcpServers": {
    "haki": {
      "url": "https://api.gethaki.space/mcp/"
    }
  }
}

This endpoint answered with an authorization challenge. The server is running, but it did not say what kind of credentials it expects.

Available tools 5

Read directly from the server with tools/list, grouped by what they act on. If a tool disappears, we record the date.

haki

haki_capture

Memorize a durable project fact: technical decision, convention, resolved error. Call at the END of a task. Never memorize secrets, tokens or ephemeral data. Consolidation is synchronous in dev (HAKI_MCP_AUTOCONSOLIDATE), so the fact is recallable immediately. Content larger than MAX_JSON_BYTES (256 KiB, same bound as POST /v1/capture) is rejected with a typed error.

haki_context

Recall the project's memory (decisions, conventions, preferences) relevant to a task. Call BEFORE planning or editing code. Returns a ready-to-inject block with dates and sources, the trace_id (inspectable via haki_inspect), and an explicit `status` ("ok"/"degraded"/"failed") — never treat a degraded or failed response as "no known facts". If the block does not hold what was needed, call this tool again with the SAME query and `exclude_ids` set to the ids already received — it serves the next page of the same ranked list. Do NOT rewrite the query with what was just read: measured on this project's bench, that finds the missing turn less often than asking unchanged.

haki_correct

Correct the memory from inside the conversation (M10). rating='incorrect' with a fact_id moves that fact to 'disputed': haki_context will never recall it again. rating='useful'/'irrelevant' logs feedback on a recall without changing the fact. Exactly one target required: fact_id (one precise fact, usually seen via haki_context/haki_inspect) OR trace_id (one whole haki_context call). Same mechanism as POST /v1/feedback (app.ledger.submit_feedback): identical effect whatever the call path. Scope resolved by API key (see _resolve_scope) or, self-hosted without a key, by HAKI_MCP_PROJECT_ID.

haki_forget

Forget the configured subject's memory in this project. mode='disable' (reversible, facts move to disabled) or 'delete' (real erasure: facts, embeddings, events, traces). Returns the erasure receipt (forget_id) and the counters of what was done.

haki_inspect

Inspect the trace of a haki_context call: which facts were included, excluded or blocked, and why (reason_code). Provenance proof of the served memory.

Endpoints

URL	Transport	State	Latency	Checked
https://api.gethaki.space/mcp/	streamable-http	needs auth	217 ms	4 min ago

Alternatives to Haki

same job, measured the same way

Vlk MemAct

by aranajhonny

Stateful memory for AI agents. Temporal awareness, error loop detection, and preventive constraints.

local only

Memharness

by las7

Bi-temporal, provenance-carrying long-term memory for AI agents, in one SQLite file.

38 installs/wk local only

Yantrikdb MCP

by yantrikos

Cognitive memory for AI agents — semantic recall, knowledge graph, and contradiction detection

1 446 installs/wk local only

Agently Brain

by agently

Company brain for AI agents — temporal knowledge graph search, exploration, and durable memory.

answering

Engram Server

by jamjet-labs

Durable memory for AI agents: fact extraction, hybrid retrieval, temporal graph. SQLite/Postgres.

local only

Memra

by usememra

Persistent memory for AI agents. EU-hosted, privacy-first, hybrid recall, contradiction detection.

7 tools answering

CP Memory

by cjhuochai

Local-first, governable long-term memory for AI coding agents

747 installs/wk local only

Codeledger

by codeledgerecf

Deterministic context selection for AI coding agents. Local memory that compounds across sessions.

435 installs/wk local only

Haki — questions

Answers built from our own checks of this server.

What can Haki do?

It exposes 5 tools, read directly from the server on our last check. Among them: haki_capture, haki_context, haki_correct, haki_forget, haki_inspect. The full list with descriptions is on this page — we take it from the server itself via tools/list, not from a README. How MCP servers expose tools in the first place →

Is Haki working right now?

We send a real MCP handshake every 15 minutes. Over the last 24 hours 92 of 92 checks got a reply (100.0%), average response time 221 ms. The bar chart above shows every period we have measured.

How do I connect Haki?

Copy the ready config from this page — we generate it for Claude Code, Claude Desktop, Codex, Cursor and VS Code, each with the file path that client actually reads. It is a remote server, so there is nothing to install — the client connects to the address. You will need your own credentials: it refuses anonymous clients.

Does Haki need an API key?

Yes. Every time we knock, Haki answers with an authorization challenge instead of its tool list — that is how we know it is running and gated rather than broken. Bring your own credentials and it will talk.

How fast is Haki?

It answers our handshake in 221 ms on average, which is faster than 62% of all working MCP servers we measure. The comparison comes from our own checks across the whole registry, every 15 minutes.

Is Haki open source?

Yes — it is published under the Apache-2.0 licence, written in Python, 2 stars on GitHub and 8 open issues. The source link is on this page, so you can read exactly what it does with your data before you connect it.