IntentKit
A self-hosted cloud agent cluster for coordinating AI agents that collaborate on tasks.
- Source repo
- crestalnetwork/intentkit
- Stars
- ★ 6.5k
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- Python
- FA score
- 35/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changes
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A team that needs a self-hosted agent service and wants collaborative agents running in the cloud rather than on high-spec local hardware.
- Not a fit if
- Users wanting a local-first personal assistant with local hardware and permissions
- Developers who expect to contribute code directly via Pull Requests
- Source review
- 35/100 · Major gaps 4 safety controls not found
What does this agent do, and when should you use it?
IntentKit is an open-source, self-hosted cloud agent cluster for managing a collaborative team of AI agents. It is positioned as cloud-native, with minimal local resource use, no local maintenance, and reliable task completion as stated goals. Its agents can call and interact with one another, giving it a collaborative execution model. The project also describes an extensible skill system plus optional Web3, blockchain, and social-media integrations. Teams can self-deploy it, extend it as a Python library, or interact with agents from external applications through built-in API endpoints.
Once deployed, IntentKit manages a collaborative team of AI agents; the README specifically says agents can call and interact with each other. External applications can interact with agents through built-in API endpoints, while developers can import and extend IntentKit as a Python Library. The project identifies an Extensible Skill System as its capability-extension point and describes optional Web3, blockchain, and social-platform integrations. The supplied material does not specify endpoint paths, request schemas, Python import names, skill interfaces, or agent output formats.
- A team that needs a self-hosted agent service and wants collaborative agents running in the cloud rather than on high-spec local hardware.
- A developer with an existing Python application who wants to import IntentKit as a Python Library and extend it with agent-cluster capabilities.
- A team with an external business application that needs to interact with deployed agents through built-in API endpoints.
- A user whose workflow requires multiple agents to call and interact with one another while completing tasks.
- A crypto or blockchain-oriented team evaluating whether the optional Web3 and blockchain integrations fit its system.
- A team evaluating a connection between its agents and social platforms through the described social-media integration.
How do you install or deploy this agent?
The README provides no copyable installation command, image-pull command, runtime version, environment variables, credential requirements, or first successful invocation. It says IntentKit can be self-deployed and points to a Deployment Guide; it also displays badges for the crestal/intentkit Docker image and the intentkit PyPI package. Required runtime, configuration, credentials, and startup steps must be confirmed from deployment documentation.
How do you use this agent?
The README describes three usage boundaries: self-deploy it to manage a collaborative agent team, import and extend it as a Python Library, or have an external application call built-in API endpoints. The supplied material provides no API URL, authentication scheme, request body, Python import, initialization code, or CLI command, so no verifiable first-call command can be given.
What are this agent's strengths and limitations?
- It explicitly supports self-hosting, which suits teams that need to control their deployment boundary.
- It uses a collaborative model in which multiple agents can call and interact with each other.
- It offers two stated integration boundaries: extension as a Python Library and interaction through built-in API endpoints.
- It identifies optional Web3, blockchain, and social-media connections as integration areas.
- The README does not provide installation commands, runtime versions, configuration details, or credential requirements, so implementation effort cannot be estimated from the supplied material.
- No API paths, authentication method, request schema, or response example is provided, requiring further validation before external-app integration.
- The material does not identify supported model providers, model configuration, or a provider-migration path.
- The project currently does not accept code contributions through Pull Requests; feature requests and bug reports are directed to GitHub Issues.
How does this agent compare with similar options?
The README contrasts IntentKit with a Local-First path, naming OpenClaw as an example: IntentKit is positioned as a cloud-native, self-hosted cloud agent cluster, while Local-First tools are described as personal assistants or experimental tools that may require expensive hardware and broad local permissions.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| IntentKit This agent | 35 · Major gaps | Self-hosted serviceFree + model costs | ★ 6.5k | 1mo ago | Python | — |
| Statewave Agent Memory Runtime | 62 · Some gaps | Self-hosted serviceFree + model costs | ★ 259 | 13d ago | Python | OpenAI API · Claude API |
| Brigade — Enterprise-grade personal intelligence | 59 · Major gaps | CLIFree + model costs | ★ 11k | 2d ago | TypeScript | ChatGPT · Codex · Claude Code · OpenAI API · Claude API |
| JoySafeter | 59 · Major gaps | Self-hosted serviceFree + model costs | ★ 315 | 6mo ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README claims 'secure by design' but lacks specific mechanisms; pyproject.toml includes sensitive services like AWS Secrets Manager but no least-privilege details; test file test_accessible_agent.py shows access control logic but no user confirmation or data flow transparency. Deductions: no user confirmation, data flow transparency, external effects control, rollback; sensitive data handling and dependency security only partially evidenced.
Evidence shows: pyproject.toml has version ranges, tests exist, but failure messages not detailed; dependency availability partially addressed (e.g., bcl override) but not verified. Deductions: failure messages not detailed, dependency availability not fully verified.
Evidence shows: README describes multiple use cases (self-deploy, Python library, API), pyproject.toml supports environments (Ollama, PDF), but capability boundaries and trigger precision not defined. Deductions: capability boundaries and trigger precision missing.
Evidence shows: README provides basic info, pyproject.toml has version and dependencies, but install steps, known limitations, changelog missing; naming stability not explicit. Deductions: install notes, known limitations, changelog absent.
Evidence shows: README claims 'out-of-the-box ready' but no output examples; marginal value in cloud-native and collaboration, but cost-benefit not quantified. Deductions: output usability not demonstrated, cost-benefit not quantified.
Evidence shows: README feature claims lack implementation evidence; tests exist but not covering all claims; facts and inferences not clearly separated. Deductions: claim traceability insufficient, cross-source corroboration limited.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- No user confirmation mechanism; agents may execute external actions automatically.
- Lack of data flow transparency; users cannot track agent access to sensitive data.
- Dependency security not fully verified; supply chain risks exist.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Release v2.6.3 Improvements - Observability traces (Langfuse) now record a more accurate per-request cost: when the provider reports the actual charge (e.g. OpenRouter) it is used directly, otherwise the cost is computed from the model catalog — and cached input tokens are now priced at their discounted rate instead of being undercounted. No user-facing changes.View on GitHub ↗ -
Release notes excerpt
Release v0.18.0 New Features - WeChat agents now accept voice, video, and file messages in addition to images. Users can send an audio note, a short video, or a document and the agent will receive the content directly instead of a "type not supported" fallback. Improvements - Expanded the catalog of selectable LLM models — the latest DeepSeek, Claude Opus, Kimi, and MiMo releases are now available when configuring agents. …View on GitHub ↗ -
Release notes excerpt
Release v0.17.60 Bug Fixes - Fixed a long-standing issue where images sent through WeChat came through as unreadable garbage and the AI refused to process them. Inbound images now decrypt correctly, so users can share photos with their agents and get a real response. Full Changelog: https://github.com/crestalnetwork/intentkit/compare/v0.17.59...v0.17.60View on GitHub ↗