Kika MCP server
online · 100% uptimeio.github.usekika/kika·v0.2.4Shared memory for client engagements: decisions with reasons, blockers, promises, what was tried.
0 stars
Reviews
Write oneNobody has reviewed Kika yet.
If you have run it, two minutes of your experience saves the next person an afternoon.
Kika tools
Tool list not cached yet. `describe` through the gateway fetches it live.
Public scan report
scanner v0.1.10 · 2026-10-02 · same rubric, same numbers if you re-run it
- –Code scanremote-only server, no package to scann/a
- Live reliabilityremote reachable in 142ms (auth required)20/20
- –Tool poisoningtools not inspected (endpoint requires auth); not countedn/a
- Auth qualityOAuth resource metadata advertised on 40115/15
- Maintenancelast push 12 days ago15/15
- Maintainer identityregistry namespace matches repository owner7/10
How Kika compares
Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.
- Safety score
- 95/100
- higher than 97% of 48,255 servers
- Uptime
- 100%
- of our recent checks; the typical remote server answers 100%
- Response time
- 242 ms
- faster than 57% of 30,031 servers
Install Kika in Claude Code, Cursor or VS Code
claude mcp add --transport http kika https://api.getkika.app/mcp
What the publisher says
From the Kika repository's README, as published. We do not edit it. Read it on GitHub
Kika for Claude Code
Kika gives a project one memory that every agent and every person on it reads. What was decided and why, what is blocked, what was promised and by when, what was already tried and dropped. None of that is in the repository, and an agent cannot infer it from the code.
One install gives you both halves:
updatememory, closeloop, so an agent can read the project's memory and add to it; and
- the MCP server — getbrief, searchmemory, remember,
having to remember to.
- the capture hook — which records what a turn settled without the agent
The hook is the part no other memory product has. A handshake that says "write things down when they matter" is advice, and a model under load ignores advice. A hook is not advice.
Install
/plugin marketplace add usekika/kika-plugin
/plugin install kika@kikaThen /mcp, choose kika, and approve in the browser.
That is the whole install. The plugin carries both halves, so there is nothing to clone, no script to run, and no second credential: the hook reads the OAuth token Claude Code already holds for this server and keeps refreshed.
Projects create themselves the first time something worth remembering is said in a repository. There is nothing to set up per project.
With a token instead
For CI, or a machine with no browser. Get one at getkika.app/app/connect:
claude mcp add --scope user --transport http kika https://api.getkika.app/mcp \
--header "Authorization: Bearer kika_mcp_…"--scope user matters: without it the server is bound to the one project you happened to be in. The hook reads that same header out of your MCP configuration, so it is still one credential.
Platform note
The OAuth path is verified on macOS, where Claude Code keeps the token in the login keychain. Elsewhere the hook looks for ~/.claude/.credentials.json first, which may or may not be where your build stores it. If capture is silent on Linux or Windows, use the --header form above.
For a teammate
They install the same way and sign in as themselves. What they do in their own repositories is their own memory.
To put them on a project of yours, add them at getkika.app/app/projects — the project, then People, then the email address they sign in with. They do not need an account first; the invitation waits and is claimed the first time they sign in.
From then on their checkout of that repository resolves to your project rather than a second copy of it: what they decide, you read, and every fact shows who recorded it and from which branch. Everyone on a project can read and write its memory; only the owner decides who else is on it.
Until you add them, a repository that already belongs to a project is refused rather than duplicated. The agent is told the project exists, that they are not on it, and that whoever set it up can add them.
What leaves your machine
When a finished turn looks like it settled something, the hook posts to https://api.getkika.app/hooks/turn:
typed (at most 4,000), keeping the end of each rather than the beginning
- the last assistant message (at most 8,000 characters) and the last message you
- the working directory, and git remote get-url origin for it
- the Claude Code session id
It does not send file contents, your repository, your environment, or any other conversation. Most turns are discarded on your machine before any request is made at all.
Shortened. The full README is on GitHub.
Nothing above is checked by us. What we check is on the safety report.
Grade history
- 2026-09-19restoreB → Ascore 95
Kika: common questions
- Is Kika MCP server safe?
- Yes, by our scan: it is graded A (95/100). Read the Kika safety report
- How do I install Kika?
- It runs remotely at api.getkika.app. Add it to Claude Code, Cursor, VS Code or Claude Desktop with the snippets above, or call it through the mcp.market gateway without installing anything.
- Does Kika need an API key?
- No key to paste: it signs you in with OAuth when your client connects.
- Is Kika maintained?
- The last commit was 45 days ago (2026-09-20). The latest release is v0.2.4.
- Is Kika up?
- 100% of our last 28 checks got an answer. We check remote servers about four times a day.
- What can I use instead of Kika?
- Servers from other publishers that do the same job: State Memory MCP server, Honcho MCP server and Tradememory Protocol MCP server.
- How does Kika compare with other MCP servers?
- safety score 95/100 (higher than 97% of 48,255 servers); uptime 100% (of our recent checks; the typical remote server answers 100%); response time 242 ms (faster than 57% of 30,031 servers). Full comparison.
Alternatives to Kika
Same job from other publishers: the closest match first, then the best rated.
State Memory MCPPersistent workflow graph server tracking tasks, decisions, spatial entities, and blockers.not reviewedEstablishedA- HonchoMemory that reasons: continual learning for stateful agents. Better context, fewer tokens.not reviewedEstablishedA
- Tradememory ProtocolTamper-evident decision audit trail and outcome-weighted memory for AI trading agents.not reviewedEstablishedB