Quick Verdict
Gread v1.0.3 is a read-oriented layer that lets AI agents search, browse, and fetch files from public GitHub repositories. It is available both as an Agent Skill and a Streamable HTTP MCP server, making it useful when an agent needs current source code rather than an answer based only on model training data.
The public endpoint is convenient, but it clones and caches repositories on Gread’s remote infrastructure. “Open source” does not mean requests stay on your machine. As of this review, the public service does not publish a data-retention policy, SLA, or availability commitment. Self-hosting gives you control over the cache, but requires infrastructure, a GitHub token, and model-provider credentials for features such as documentation-repository detection.
Best For
- Developers who routinely inspect open-source dependencies and examples.
- Users adding read-only public repository context to Cursor or another coding agent.
- Teams able to operate a Bun/TypeScript service and manage provider credentials.
- MCP users who need source discovery without broad repository write permissions.
- Not a fit for private-repository access, contractual retention requirements, or an unqualified production dependency on a free public endpoint.
Key Features
- Skill and MCP access for coding agents and general MCP clients.
- Repository discovery by name, description, and topic.
- Tree browsing before file retrieval, reducing unnecessary context.
- Repository search backed by server-side Git operations after cloning.
- Exact-path reads for one or more source files.
- Documentation-repository detection, which is model-assisted and should be verified.
- Apache-2.0 self-hosting for teams that want to control storage and operations.
Use Cases
A practical workflow is to find a repository, inspect its tree, search for a symbol, and fetch only the implementation and tests that matter. This is useful for checking how an SDK actually handles an option, tracing a type definition, or grounding an agent before it proposes an integration.
Gread is narrower than GitHub MCP Server. It focuses on reading public source rather than managing issues, pull requests, or repository changes. That smaller permission surface is useful, but users should still verify commit references, paths, and license terms on GitHub.
Pricing
The v1.0.3 source is licensed under Apache 2.0, and the hosted Skill/MCP is currently presented as free. There is no published service-level guarantee or contract stating that current limits and availability will remain unchanged.
Self-hosting removes software license fees, not operating costs. Budget for storage, network traffic, GitHub API use, maintenance, and any model-provider calls. Tokens and provider keys should have minimal scopes, rotation, and redacted logging.
Pros
- Clear progressive workflow: discover, browse, search, then read.
- Works as both an Agent Skill and MCP server.
- Open source under Apache 2.0 and can be self-hosted.
- Remote caching can reduce repeated repository downloads and search calls.
- Read-oriented scope is easier to govern than a full GitHub automation tool.
Cons
- The public service is a remote cache, not a local-only privacy solution.
- No published data policy, SLA, support window, or capacity commitment.
- Public GitHub repositories are the supported scope; private code is not.
- First access and large repositories consume time, bandwidth, and disk.
- Model-assisted docs-repository matching can be wrong or incomplete.
Alternatives
| Tool | Best for | Main difference | Caveat |
|---|---|---|---|
| Context7 | Version-aware library documentation | More documentation-focused | Different source coverage and service limits |
| GitHub MCP Server | Issues, PRs, and repository operations | Much broader GitHub object support | Write-capable tools need strict approval |
| Sourcebot | Self-hosted code search with a UI | Richer indexing and search product | Heavier deployment and index management |
| Cursor | Daily work inside an IDE | Full editing and project workflow | Not a standalone reader for every public repo |
FAQ
Is Gread free?
The source and current public endpoint are offered free of charge. Self-hosting still incurs infrastructure, GitHub, and model-provider costs, and the hosted endpoint has no published SLA.
Can Gread read private repositories?
Its published scope is public GitHub repositories. Do not submit private code or internal repository credentials to the public service.
Where is the repository cache stored?
With the public endpoint, repositories are cloned and cached on the service operator’s infrastructure. With self-hosting, the cache is on infrastructure you control.
What does self-hosting require?
In addition to the runtime and persistent storage, expect to configure a GitHub token. Model-provider configuration and credentials are needed for model-assisted features such as docs-repository detection.
Does Gread replace GitHub MCP Server?
No. Gread is optimized for public-source reading. GitHub MCP Server covers a broader set of GitHub objects and operations.
Should production systems depend on api.gread.dev?
Use it for evaluation and non-critical workflows first. Without a published SLA, retention policy, or capacity commitment, critical systems should self-host or have a fallback.
Bottom Line
Gread is useful because it gives agents a disciplined way to inspect public source without dumping entire repositories into context. The hosted MCP is convenient for individual research. Teams needing predictable operations, auditable caching, or explicit deletion rules should self-host and test their own failure and storage boundaries.