flyte-mcp is an MCP server that gives AI coding assistants knowledge and tools to author, migrate, and run Flyte V2 code.
Explore the Flyte V2 API: List public symbols, look up details (signature, docstring, module path), and search across the API by keyword.
Get oriented with Flyte: Access the SDK overview, feature list, install guide, and version/count info.
Find canonical code examples: Browse example themes, retrieve full pattern code and READMEs, and search examples by natural-language use case.
Discover and select plugins: List available plugins (with optional category filtering), get full plugin details, and get suggestions based on a natural-language need (e.g. "run Spark" or "connect Snowflake").
Migrate V1 code to V2: Automatically rewrite flytekit V1 Python code into flyte-sdk V2 syntax, with a list of applied changes and manual-review warnings.
Execute tasks on a Flyte cluster: Run a task from a Python file, check execution status by run ID, and list recent runs from the configured cluster.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@flyte-mcp show me how to cache a task with GPU resources"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
claude mcp add flyte -- uvx flyte-mcp{
"mcpServers": {
"flyte": {
"command": "uvx",
"args": ["flyte-mcp"]
}
}
}Add this to ~/.claude.json, ~/.cursor/mcp.json, or your client's config file.
Ask any AI assistant "write a Flyte V2 task with caching and 4 GPUs" and you get confidently wrong code: V1 imports, invented decorators, hallucinated resource kwargs. The assistant has no reliable channel into the Flyte ecosystem, so it fills the gap with training-data guesses.
flyte-mcp is that channel. It ships a versioned knowledge pack built directly from the flyte-sdk source tree and the Flyte Plugin Registry, plus a thin runtime bridge for executing tasks when a cluster is configured. The assistant stops guessing and starts answering.
Related MCP server: Snippy
Capability | Tools |
Learn the V2 API |
|
Find canonical examples |
|
Pick the right plugin |
|
Port V1 code to V2 |
|
Get oriented |
|
Run on a cluster |
|
All tools are pure Python, stdio transport, zero network calls unless you explicitly use the runtime bridge.
How do I cache a task and invalidate on input change?
Show me a distributed PyTorch training example with A100s.
Which Flyte plugin do I use for Snowflake, and what's the import?
Migrate this flytekit V1 workflow to V2.
What's the signature of TaskEnvironment?
Your assistant picks the right tools and assembles accurate answers.
Contributors and release automation can regenerate the pack from source:
python scripts/build_knowledge.py \
--sdk-path /path/to/flyte-sdk \
--registry /path/to/flyte-plugin-registry/src/data/plugins.json \
--out src/flyte_mcp/data/flyte-v2-knowledge.jsonSources used:
flyte-sdk/src/flyte/__init__.py — public symbols via AST
flyte-sdk/examples/* — canonical patterns by theme
flyte-sdk/README.md, FEATURES.md, CONTRIBUTING.md — meta docs
flyte-plugin-registry — curated plugin catalog
flytesnacks is intentionally excluded: V2 consolidated examples in-tree.
wherobots/flyte-mcp — runtime-only. Discovers and executes tasks on a deployed Flyte instance via API key. Complementary, not competing.
unionai/claude-agents-public — Claude Code custom agents (system prompts, not an MCP server). Compose freely.
This project focuses on authoring: the moment a developer types a prompt asking about Flyte.
GitHub Action to auto-rebuild the knowledge pack on every flyte-sdk release
Listing in the official MCP Registry
Local semantic search via small sentence-transformer model
AST-based migration codemod (replacing the current regex pass)
Log streaming tool (get_flyte_execution_logs) with tail support
Apache-2.0 — same license as Flyte itself.
Independent community project. Not officially affiliated with or endorsed by Flyte or Union.ai. The Flyte name and logo are trademarks of their respective owners.
Find example themes best matching a natural-language use case.
Searches READMEs and file docstrings across all example themes.
Inputs: use_case (required), limit. Returns: result
Fetch current status of a Flyte run by id.
Inputs: run_id (required).
Return the flyte-sdk FEATURES.md - concise list of V2 capabilities.
Inputs: no named parameters. Returns: result
Return CLI help output and install instructions parsed from flyte-sdk.
Inputs: no named parameters.
Return the flyte-sdk README - high-level description of Flyte V2.
Inputs: no named parameters. Returns: result
Return README and example code for a theme (e.g. 'caching', 'genai', 'accelerators').
If file is given, return just that file's snippet. Otherwise README + first file.
Inputs: theme (required), file.
Full detail of a Flyte plugin by slug: package, modules, versions, install command.
Inputs: slug (required).
Full detail of one Flyte V2 symbol: signature, params, docstring, module path.
Accepts "flyte.TaskEnvironment" or just "TaskEnvironment".
Inputs: name (required).
Flyte SDK version this knowledge pack was built from, plus counts.
Inputs: no named parameters.
Canonical example themes from flyte-sdk/examples/.
Inputs: no named parameters. Returns: result
List Flyte plugins from the registry. Optional category filter.
Inputs: category. Returns: result
List recent runs from the configured Flyte cluster.
Inputs: limit. Returns: result
List public Flyte V2 API symbols. Optional kind filter: function, class, constant.
Inputs: kind. Returns: result
Rewrite flytekit V1 Python code into flyte-sdk V2 syntax.
Returns the transformed code plus a list of applied changes and warnings about patterns that need manual review.
Inputs: code (required).
Execute a task on the configured Flyte cluster. Requires flyte-sdk installed.
module_path: absolute path to the .py file defining the task task_name: Python name of the task function within that file inputs: dict of input kwargs
Inputs: task_name (required), module_path (required), inputs.
Keyword search over Flyte V2 symbol names, signatures, docstrings.
Inputs: query (required), limit. Returns: result
Given a natural-language need (e.g. 'run Spark', 'connect Snowflake'), suggest plugins.
Prefers V2-native entries (sdk='flyte-sdk') over legacy V1 duplicates.
Inputs: need (required), limit. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 17 tools
Each tool targets a distinct aspect of Flyte: documentation retrieval, example exploration, plugin details, symbol lookup, execution status, task execution, code migration, and search. There is no overlap in purpose.
Tool names follow a consistent imperative verb_noun pattern with underscores (e.g., get_flyte_, list_flyte_, find_flyte_, run_flyte_task). The naming is predictable and uniform across the set.
17 tools is slightly above the typical 3-15 range, but each tool serves a unique and necessary function for the Flyte ecosystem. The count is well-scoped and not excessive.
The tools cover the primary use cases for interacting with Flyte: documentation, examples, symbols, plugins, runs, task execution, and migration. Missing features like workflow creation are outside the intended scope, so there are no significant gaps.
Versioned documentation registry and semantic search for AI tools and coding assistants.
Provide your AI coding tools with token-efficient access to up-to-date technical documentation for…
Ingest, manage, and retrieve documents for RAG-powered AI applications
Governed, auditable knowledge your team curates for its AI assistants, self-hostable
