Skip to content

Discoverability and positioning

SchemaRouter should be discoverable without pretending to be a general agent framework, MCP server, or model router.

Canonical positioning

Use this sentence as the default long-form description:

SchemaRouter is a typed capability routing and governed execution layer for AI agents across APIs, tools, and data systems.

Supporting variants may be shorter, but they must preserve the same product boundary.

Short variant

Typed capability routing and governed execution for AI agents.

Ecosystem variant

Put APIs, MCP tools, SDKs, and data systems behind one typed capability boundary instead of dumping the entire capability catalog into model context.

What it is / is not

SchemaRouter is:

  • a typed tool/capability registry;
  • a capability-retrieval layer;
  • a schema-aware planning and execution boundary;
  • an MCP/OpenAPI/Python/tool-framework integration layer;
  • a validation, projection, policy, health, and schema-lifecycle boundary.

SchemaRouter is not:

  • a general agent framework;
  • an LLM provider gateway;
  • a model router;
  • an MCP server catalog;
  • a document retriever or final-answer generator.

This distinction matters for both technical accuracy and search quality. A visitor searching for "tool routing" should not be led to believe that SchemaRouter routes between LLM models.

Search vocabulary

Use these terms naturally where they describe real capabilities:

  • agent tool routing;
  • typed tool registry;
  • capability retrieval;
  • schema-aware execution;
  • MCP tools;
  • OpenAPI tools;
  • JSON Schema validation;
  • LangChain tools;
  • LangGraph integration;
  • LlamaIndex tools;
  • RAG tool execution;
  • provider/access fallback;
  • field projection.

Do not repeat terms solely for ranking.

Surface alignment

The project should keep the following aligned for every stable release:

Surface Requirement
GitHub description one-sentence typed capability/execution positioning
GitHub homepage published documentation site
GitHub topics protocols/frameworks actually supported
README first screen problem statement + canonical positioning + install
PyPI summary same product category and boundary
PyPI keywords real protocols/frameworks/search terms only
Docs home current stable version and same product boundary
Release notes stable product claims separated from research claims

Current topics already cover most of the intended surface. When editing repository settings, retain the accurate existing topics and add the missing high-signal terms when useful:

  • langgraph;
  • capability-retrieval;
  • schema-aware-execution.

Do not add unrelated high-volume topics.

The repository homepage should remain the published docs URL. Social preview is handled by the release checklist.

External listing opportunities

Listing submissions are gated by the destination's own rules and by SchemaRouter's evidence.

Destination Fit Current status
Christian-Sidak/awesome-mcp-tools MCP frameworks/tools Defer until there is demonstrated external/community value; its rules reject shallow self-promotion.
kaushikb11/awesome-llm-agents Agent Infrastructure Not yet eligible: its current policy requires at least 25 stars unless published by a recognized organization/research lab.
awesome-llms-labs/awesome-ai-agents Agent infrastructure/ecosystem Defer until genuine external usage/influence is documented.
Model-routing awesome lists Routes between LLM models Do not submit; SchemaRouter is a tool/capability router, not a model router.
MCP server-only catalogs MCP servers Do not submit; SchemaRouter consumes/integrates MCP capability sources but is not itself an MCP server directory.

LangChain/LlamaIndex distribution is tracked separately in issue #10 because upstream package/listing policy can change independently of general discoverability work.

Release discoverability checklist

Before a stable release:

  • [ ] README, docs home, PyPI summary, and release notes name the same stable version.
  • [ ] Canonical positioning still matches the actual product boundary.
  • [ ] New protocol/framework names are added only after first-class support exists.
  • [ ] Deprecated/removed integrations are removed from search metadata.
  • [ ] PyPI keywords and project URLs still resolve to maintained surfaces.
  • [ ] GitHub description, topics, homepage, and social preview are reviewed.
  • [ ] Example gallery contains at least one runnable path for each advertised major integration.
  • [ ] Benchmark/research claims link to reproducible evidence and remain separate from stable product claims.
  • [ ] External listing eligibility is re-checked before submission; never assume an old contribution policy still applies.

Submission rule

Do not submit SchemaRouter to an external directory merely to gain a backlink. Submit only when:

  1. the category accurately describes the project;
  2. the repository satisfies that directory's current objective requirements;
  3. the linked quickstart/docs work;
  4. the description avoids unsupported performance claims;
  5. external usage/evidence is sufficient for lists that require community value.