What it does
5 features
NanoClaw · Self-hosted personal agent
Product and access
NanoClaw is listed as a separate product profile.
Core features
5 features from provider sources · checked on 16 October 2026. Availability and limits are listed below.
Usable AI models
Open choice
By default NanoClaw runs Claude through the Claude Agent SDK. Bundled skills add Codex, Cursor, OpenCode with OpenRouter, Google or DeepSeek, and local models through Ollama, set per agent group. The README does not list arbitrary OpenAI-compatible endpoints as a separate route.
Model families named
- Anthropic (Claude)
- OpenAI (GPT)
- Google (Gemini)
- DeepSeek
- Other models can be connected freely
- Model choice
- You choose
- Your own API key
- Yes
- Router or custom endpoint
- Yes
- Local models
- Yes
The rating covers the generally available main access. Documented · Sources accessed on
Does it fit my task?
2 practical tasks · 3 requirements and limits
A good fit if …
- you can run Docker, Node.js and pnpm on your own machine or server
- you want agents isolated in containers without holding raw credentials
- you want to connect channels such as WhatsApp, Telegram or Slack yourself through skills
Less suitable if …
- you expect a monitoring dashboard, as the repository lists none
- you do not want to run your own infrastructure, as you install and operate Docker, Node.js and pnpm yourself
Editorial assessment based on provider documentation. Not a product test.
Practical tasks
Editorial examples based on documented features. No product tests performed.
Request a briefing through messaging
- Your input
- A topic and three approved web sources.
- Possible result
- A summary with sources in the connected channel.
- What to check
- Check content, permissions and actions taken.
Separate two workspaces
- Your input
- Two agent groups with different sample files.
- Possible result
- Separate tasks with traceable file access.
- What to check
- Record time spent, corrections and costs.
Request a briefing through messaging.
Your possible prompt
Compare three providers for [task]. Use their official documentation. Show differences, sources, and open questions in a table.
Adjust the task, data and permissions.
Requirements and limits
Can I use it?
With requirements
Europe & regional availability
Installation instructions and model prerequisites reviewed. The assessment covers the local operating path described; complete country and feature availability of all connected services remains open.
Reviewed scope: Local software and documented self-hosting; external services assessed separately
Reviewed on:
With requirements
Self-hosting with own model access
NanoClaw installs its own container runtime and uses Anthropic credentials by default. Local Ollama models are documented as an alternative extension; check selected models and chat channels separately.
Region, reason and source
Region covered: Europe: own computer or server; model services are country-specific
Documented reason: Technical assessment based on installation and provider selection. These sources do not establish a legal cause for regional restrictions.
Accounts & permissions
Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.
Configuration covered: Self-hosted NanoClaw (open source, MIT) with Docker containers and messenger channels
Sources reviewed:
Show all 3 points
Documented with conditions
Admin approval
Approvals run through cards sent to admins by direct message, for example for credentials, self-modification and unknown chats. If no approver is reachable, the request is denied automatically.
Evidence
Memory and conversation continuity
2 of 6 criteria documented
Shared agent memory and shared conversation can be configured separately.
NanoClaw V2: agent groups and channel mapping
Sources accessed:
Documentation by criterion
StorageDocumented
Storage and scope
Agent groups own a workspace, CLAUDE.md and persistent memory under groups/; separate groups have their own directories and containers.
SelectionPartly documented
Selection and timing
Agents can save information when asked to remember. Exact automatic selection rules and write timing are unspecified here.
Edit & deletePartly documented
Inspect, edit and delete
Local workspace and CLAUDE.md files are accessible; a dedicated interface for individual memory facts is undocumented.
Access boundariesDocumented
Agent and channel boundaries
Channels can share session and memory, share only memory, or belong to fully separate agent groups.
RetrievalPartly documented
Retrieval
Saved group memory can be recalled in other associated sessions; the search algorithm is not explained here.
New conversationsPartly documented
Session reset and compaction
Separate conversations of the same agent group retain shared memory. Specific reset and compaction behavior remains unspecified in these sources.
Cost and effort
Prices as of 16 October 2026
Prices and included usage
Pricing checked
Open source; model and operating costs are separate.
- Total cost depends on access, usage and configuration.
Setup
Set up Docker, Node.js and model access.
Operation & oversight
Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.
Deployment covered: NanoClaw on your own machine or server, agent groups in containers, scheduled tasks via the CLI
Sources reviewed:
Editorial assessment
You run Docker, updates and troubleshooting yourself, though approvals and permissions are documented in detail. You need to check scheduled-task failures actively via the CLI.
Show all 6 points
Open in reviewed sources
Sign-ins
The reviewed pages do not say how long channel connections remain valid. They only state that the container does not start if the credential vault is unreachable.
EvidenceDocumented
Starting further runs
Tasks run once by timestamp or recurrently by cron in the installation's timezone, optionally with a script gate. An ungated recurring task that would fire more than four times in 24 hours is rejected.
EvidenceDocumented with conditions
Human intervention
Approvals exist for specific actions (vault credentials, self-modification, unknown chats, optionally agent-to-agent messages). Not every action is confirmed individually; without a policy, agent-to-agent messages flow freely.
EvidenceDocumented with conditions
Resuming after failures
Failed recurring runs are delayed in steps from 2 to 60 minutes; after eight consecutive failures the series pauses until you fix and resume it. Retrying a single run is not described. If a setup step fails, Claude Code is invoked automatically to diagnose and resume from there.
Evidence · Evidence
Evidence and open questions
1 source · 2 open points
Open points
The reviewed sources contain no reliable information on these points. A gap says nothing about the product.
Requirements and limits: not yet confirmed
Sources and evidence
Provider sources reviewed; examples are editorial. No hands-on tests.
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
Provider
Publishes NanoClaw and the linked product information.
- Operator
- NanoCo
- Organization
- Open-source project
- Website
- github.com
Changes to this profile 1
Added NanoClaw with capabilities, sources, pricing and open European availability questions.
These notes describe editorial changes. Source and pricing review dates appear in their respective sections.