Self-hosted MCP server for AI agent memory — 14 core tools, profile-based tiering (86+ available). 88.2% LoCoMo. Works with Claude, Cursor, Windsurf.
- ✓Actively maintained (<30d)
- ✓Clear description
- ✓Topics declared
- ✓Documented (README)
- !Licence file present but not machine-readable
claude mcp add dakera-mcp -- npx -y @dakera-ai/dakera-mcp{
"mcpServers": {
"dakera-mcp": {
"command": "npx",
"args": ["-y", "@dakera-ai/dakera-mcp"],
"env": {
"DAKERA_ROOT_API_KEY": "<dakera_root_api_key>",
"DAKERA_API_KEY": "<dakera_api_key>"
}
}
}
}DAKERA_ROOT_API_KEYDAKERA_API_KEYMCP Servers overview
# ⚡ dakera-mcp
[](https://github.com/Dakera-AI/dakera-mcp/actions/workflows/ci.yml) [](https://crates.io/crates/dakera-mcp) [](https://www.npmjs.com/package/@dakera-ai/dakera-mcp) [](https://crates.io/crates/dakera-mcp) [](LICENSE) [](https://dakera.ai/benchmark) [](https://glama.ai/mcp/servers/Dakera-AI/dakera-mcp) [](https://dakera.ai/docs) [](https://dakera.ai) [](https://dakera.ai/playground)
MCP server for Dakera AI. Gives any MCP-compatible AI agent persistent, queryable memory — with smart token management built in.
Works with Claude, Claude Code, and any MCP-compatible framework.
Part of [Dakera AI](https://dakera.ai) — the memory engine for AI agents.
> The Dakera memory engine scores **88.2% Recall@20 on LoCoMo** (1,536 evaluated questions · LLM-judged retrieval recall) — [benchmark details](https://dakera.ai/benchmark)
---
## Architecture: 14 core tools + on-demand discovery
Starting every agent session with 60+ tool schemas wastes ~15K tokens before you write a single message. dakera-mcp solves this with **hybrid tool exposure**:
- **14 tools loaded by default** — the 12 highest-frequency memory operations + 2 meta-discovery tools
- **On-demand expansion** — use `dakera_discover_tools` and `dakera_load_tools` to fetch additional tool schemas only when you need them
### Default tool set (core profile)
| Tool | Purpose |
|---|---|
| `dakera_store` | Store a memory with importance, tags, and type |
| `dakera_recall` | Semantic recall by query text |
| `dakera_search` | Advanced memory search with tag/type filters |
| `dakera_session_start` | Start a session to group related memories |
| `dakera_session_end` | End a session with optional summary |
| `dakera_batch_recall` | Bulk filter-based recall (by tags, importance, time) |
| `dakera_forget` | Delete specific memories by ID |
| `dakera_hybrid_search` | Combined vector + BM25 search |
| `dakera_fulltext_search` | BM25 full-text search |
| `dakera_knowledge_graph` | Build a knowledge graph from a seed memory |
| `dakera_extract` | Extract entities and structure from free-form text |
| `dakera_batch_forget` | Bulk delete by tags, type, or time range |
| `dakera_discover_tools` | Search the full tool catalog by keyword or tier |
| `dakera_load_tools` | Load full schemas for specific tools on demand |
### Profiles & token cost
| Profile | Tools | ~Tokens | How to enable |
|---|---|---|---|
| **core** | 14 | ~3,350 | Default — always loaded |
| **admin** | 34 | ~6,700 | `DAKERA_MCP_PROFILE=admin` |
| **power** | 79 | ~16,600 | `DAKERA_MCP_PROFILE=power` |
| **all** | 99 | ~19,950 | `DAKERA_MCP_PROFILE=all` |
Token figures are estimates (JSON bytes / 3). The attachment tools below count in `power` and `all`,
but only appear while the connected server has the feature on (see [Dakera v0.12](#dakera-v012)).
### Accessing additional tools
```
# In your agent: discover what's available
dakera_discover_tools(tier="power")
→ returns names + descriptions, no schemas loaded
# Load schemas for the tools you want
dakera_load_tools(tools=["dakera_consolidate", "dakera_agent_stats"])
→ returns full inputSchema for each tool
```
### Profile selection
The profile controls which tools appear in `tools/list`. Three ways to set it:
**1. Per-request** (in `tools/list` params):
```json
{"profile": "power"}
```
**2. Environment variable** (applies to all requests):
```bash
DAKERA_MCP_PROFILE=power
```
**3. Default**: `core` (14 tools, ~3,350 tokens)
---
## Dakera v0.12
dakera-mcp 0.11 works against **Dakera v0.11.108 and v0.12.0** servers. Every v0.11 tool keeps
its name and arguments; the v0.12 additions are optional arguments that are sent only when you
supply them, and tools that call v0.12 routes.
| Dakera server | dakera-mcp 0.11 |
|---|---|
| v0.12.0 | every tool; the attachment tools while `DAKERA_ATTACHMENTS` (and `DAKERA_VISION` for images) is on |
| v0.11.108 | every v0.11 tool unchanged; the attachment tools, `dakera_encryption_status` and `dakera_embed_migration_status` are not listed (a direct call says they need v0.12); `dakera_encryption_rotate_key` needs `new_key` |
| older | not tested |
### What is new
| Tool | Tier | Needs | What it does |
|---|---|---|---|
| `dakera_capabilities` | power | v0.12 | `GET /v1/capabilities`: active model, search mode, scoring strategy, accepted `lang` values, which opt-in features are on. On a v0.11 server it answers `capabilities_available: false` |
| `dakera_health` | power | any | `GET /health`: status and version; on v0.12 also `degraded`, `config_warnings`, `embed_migration` |
| `dakera_wake_up` | power | any | an agent's startup context in one call: its top memories by importance x recency, no query, no embedding |
| `dakera_embed_migration_status` | admin | v0.12 | progress of the one-time background re-embed after the upgrade |
| `dakera_encryption_status` | admin | v0.12 | the encryption keyring and the background re-seal (never key material) |
| `dakera_encryption_rotate_key` | admin | v0.11+ | `new_key` is optional on v0.12 once encryption is on (the server generates a key; with encryption off, `new_key` turns it on); `wait_secs` (at most 20); `namespace` rotates one namespace |
| `dakera_attachment_upload` / `_list` / `_download` / `_delete` | power | v0.12 + `DAKERA_ATTACHMENTS` | files (or text) a memory can reference with `dakera_store` `attachment_ref` |
| `dakera_attachment_transcribe` | power | v0.12 + `DAKERA_ATTACHMENTS` | WAV speech to text in any language the model knows (`lang` forces one) into a memory; a background job, `wait_seconds` (at most 45) waits for it |
| `dakera_attachment_index_image` | power | v0.12 + `DAKERA_ATTACHMENTS` + `DAKERA_VISION` | PNG page as a visual memory (use an agent dedicated to images: the visual lane stores page vectors) |
| `dakera_attachment_job` | power | v0.12 + `DAKERA_ATTACHMENTS` | status of a transcription / index job |
Per-request **`lang`** (`en`, `de`, `fr`, `es`, `it`, `pt`, `nl`; v0.12) is accepted by `dakera_store`,
`dakera_recall`, `dakera_recall_associated`, `dakera_search`, `dakera_memory_update`, `dakera_extract`,
`dakera_auto_tag` and the attachment jobs;
`dakera_store` also takes `attachment_ref` (`sha256:<hex>` of an attachment in the agent's own namespace,
`_dakera_agent_<agent_id>`). Neither is sent unless given, so the same calls work on a v0.11.108 server.
Other optional arguments (all servers): `ttl_seconds` and `metadata` on `dakera_store`; `tags`,
`memory_type` and `session_id` filters on `dakera_recall`; `limit` on `dakera_batch_recall`;
`limit` / `offset` on `dakera_session_list`, `dakera_session_memories` and `dakera_agent_sessions`
(the server pages at 50); `memory_type` on `dakera_knowledge_deduplicate`; `dedup_on_store` /
`dedup_threshold` on `dakera_memory_policy_set`.
### Features the server has off are left out
The opt-in features (attachments, speech to text, image indexing) are off by default on the server.
dakera-mcp asks `GET /v1/capabilities` (once a minute, 3 s timeout) before it lists tools:
* `attachments` off, or a server without `/v1/capabilities` (v0.11): the `dakera_attachment_*` tools are not
listed and not returned by `dakera_discover_tools`; a direct call answers with the variable to set
(`DAKERA_ATTACHMENTS=1`) and makes no request.
* `vision` off: `dakera_attachment_index_image` is left out (`DAKERA_VISION=1` turns it on).
* A server without `/v1/capabilities` (v0.11) also hides `dakera_encryption_status` and
`dakera_embed_migration_status`, whose routes are new in v0.12.
* The server cannot be asked (down, starting, key refused): nothing is hidden (asked again after 10 s).
The default `core` profile has no opt-in tools, so it never makes that request.
### Errors
Error answers keep the server's text and add a `Hint:` line for the v0.12 cases: a key **pinned to
namespaces** gets `403` on node-wide `/admin` routes (backups, encryption, quotas, config); backup
download, upload and restore need `super_admin`; `413` (body over a limit, or a hard quota), `501`
(feature off), `503` (`Retry-After`, which the retry logic now honours, up to 8 s) and `429` (rate
limit). A route the server lacks (an older server) and a v0.11 rotation without `new_key` get a hint
too. A request that timed out is retried only when it is safe to repeat (GET, PUT, DELETE): a store,
an import or a key rotation is never sent twice.
---
## Run Dakera
The MCP server connects to a Dakera memory server. You need one running first:
```bash
docker run -d \
--name dakera \
-p 3000:3000 \
-e DAKERA_ROOT_API_KEY=dk-mykey \
ghcr.io/dakera-ai/dakera:latest
```
For persistent storage (recommended):
```bash
curl -sSfL https://raw.githubusercontent.com/Dakera-AI/dakera-deploy/main/docker-compose.yml \
-o docker-compose.yml
DAKERA_API_KEY=dk-mykey docker compose up -d
curl http://localhost:3000/health # → {"status":"ok"}
```
Full deployment guide (Docker Compose, Kubernetes, Helm): [dakera-deploy](https://github.com/Dakera-AI/dakera-deploy)
---
## Install
### npm / npx (Node.js 18+)
```bash
# Global install
npm install -g @dakera-ai/dakera-mcp
# Or run directly without installing
npx @dakera-ai/dakera-mcp
```
### Homebrew (macOS / Linux)
```bash
brew install dakerWhat people ask about dakera-mcp
What is Dakera-AI/dakera-mcp?
+
Dakera-AI/dakera-mcp is mcp servers for the Claude AI ecosystem. Self-hosted MCP server for AI agent memory — 14 core tools, profile-based tiering (86+ available). 88.2% LoCoMo. Works with Claude, Cursor, Windsurf. It has 8 GitHub stars and its last recorded update is dated 2026-10-01.
How do I install dakera-mcp?
+
You can install dakera-mcp by cloning the repository (https://github.com/Dakera-AI/dakera-mcp) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is Dakera-AI/dakera-mcp safe to use?
+
Our security agent has analyzed Dakera-AI/dakera-mcp and assigned a Trust Score of 80/100 (tier: Trusted). See the full breakdown of passed checks and flags on this page.
Who maintains Dakera-AI/dakera-mcp?
+
Dakera-AI/dakera-mcp is maintained by Dakera-AI. The last recorded GitHub activity is dated 2026-10-01, with 3 open issues.
Are there alternatives to dakera-mcp?
+
Yes. On ClaudeWave you can browse similar mcp servers at /categories/mcp, sorted by popularity or recent activity.
Deploy dakera-mcp to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/dakera-ai-dakera-mcp)<a href="https://claudewave.com/repo/dakera-ai-dakera-mcp"><img src="https://claudewave.com/api/badge/dakera-ai-dakera-mcp" alt="Featured on ClaudeWave: Dakera-AI/dakera-mcp" width="320" height="64" /></a>More MCP Servers
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
An open-source AI agent that brings the power of Gemini directly into your terminal.
Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface
🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling_dev
The fastest path to AI-powered full stack observability, even for lean teams.