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., "@rss-publisher-mcp Publish a new article entry titled 'Hello World' to the RSS feed"
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.
Deterministic, generic RSS publishing engine for bigbatmanorg/rss-publisher-agent and other MCP/Python/CLI clients.
Repository target: https://github.com/bigbatmanorg/rss-publisher-mcp
Generic FeedEntry model (article, note, status, release, incident, media, etc.)
Immutable RSS GUIDs and explicit lifecycle (published, corrected, retracted, archived)
SQLite canonical state
Deterministic RSS 2.0 rendering with Content Module, Media RSS, Dublin Core, Atom and optional rp:* metadata
Stable local permalinks and public URL normalization from RSS_PUBLIC_BASE_URL
Publisher-owned content-addressed media/file assets
Optional OpenAI-compatible embeddings for candidate retrieval among currently active feed entries only
Strict no-op detection
Atomic static output writes and rebuild/recovery from SQLite
MCP tools, Python API, CLI, and a small HTTP upload/discovery API
It does not do research, editorial reasoning, Docker Agent orchestration, A2A, OpenAI chat serving, Caddy, or external authentication workflows. Those belong to rss-publisher-agent.
Related MCP server: Clearon Content Archive MCP
uv tool install git+https://github.com/bigbatmanorg/rss-publisher-mcp.gitDuring development:
uv sync --extra dev
uv run pytestcp .env.example .env
set -a; . ./.env; set +a
uv run rss-publisher-mcpFor the all-in-one appliance the same package runs a private loopback Streamable HTTP MCP:
RSS_MCP_TRANSPORT=streamable-http uv run rss-publisher-mcpDefault loopback endpoint: http://127.0.0.1:8766/mcp.
uv run rss-publisher-apiUpload:
curl -F 'file=@picture.webp' http://127.0.0.1:8765/api/v1/assetsResponse includes a stable content-addressed asset_id and the public URL derived from RSS_PUBLIC_BASE_URL.
A new entry gets urn:uuid:<uuid> when no ID is supplied. Once created, the ID cannot be changed.
update_entry refuses to revive an archived entry. Revival is explicit via republish_entry.
Submitting the same semantic state does not change updated_at, feed revision, feed XML, embedding, or public file mtime.
FeedConfig.max_items defines the active RSS window. Only those emitted entries are kept in the active embedding index and returned by fuzzy matching. Historical state remains in SQLite so GUIDs are never reused.
Embeddings retrieve candidates; they never automatically merge entries. Exact GUIDs and continuity keys are authoritative. The agent decides whether a fuzzy candidate is actually the same continuing item.
rss-publisher validate
rss-publisher audit
rss-publisher rebuild
rss-publisher export-state --output state.jsonThe suite covers:
immutable GUIDs
create/update/no-op behavior
explicit archive/republish/retraction
deterministic byte-identical rendering
retention/active index semantics
content sanitization
RSS namespace mappings
content-addressed asset upload and deduplication
rejection of active HTML assets
optional semantic matching with a fake embedding provider
optimistic revision conflicts
optional bearer auth on the upload API
FreshRSS private-network compatibility warnings
Run:
uv run pytest --cov=rss_publisher --cov-report=term-missing/data/publisher.db authoritative state
/srv/public/feed.xml generated RSS
/srv/public/index.html generated landing page
/srv/public/entries/* generated permalinks
/srv/public/media/* content-addressed media
/srv/public/files/* content-addressed generic filesThe public tree is a projection. It can be rebuilt from SQLite.
Uploaded HTML/XHTML/SVG active content is rejected by default.
Article HTML is sanitized through nh3.
Public media/files are content-addressed and immutable.
HTTP upload authentication is optional (RSS_AUTH_MODE=none|bearer).
The final appliance keeps this MCP private; remote clients interact with the RSS Publisher Agent instead.
Return interoperability warnings, including private URL/FreshRSS gotchas.
Inputs: no named parameters. Returns: object
Update feed/channel configuration.
Inputs: patch (required), expected_revision. Returns: object
Publish a visible correction while preserving entry identity.
Inputs: entry_id (required), correction_note (required), patch. Returns: object
Create a definitely-new entry. Generates a permanent urn:uuid ID if omitted.
Inputs: entry (required), expected_revision. Returns: object
Find semantic candidates only among currently active/published entries. Similarity is candidate retrieval, not authority to merge.
Inputs: text, entry, limit. Returns: object
Return metadata/public URL for an already-uploaded asset. Binary upload itself is performed over the HTTP asset API.
Inputs: asset_id (required). Returns: result
Return the complete canonical FeedEntry by immutable ID.
Inputs: entry_id (required). Returns: result
Return feed configuration, current revision, and publication URL.
Inputs: no named parameters. Returns: object
Return compact summaries of entries currently emitted into feed.xml.
Inputs: no named parameters. Returns: object
Apply multiple creates/updates in one publisher transaction and rebuild once.
Inputs: operations (required), expected_revision. Returns: object
Regenerate deterministic public artifacts from authoritative SQLite state.
Inputs: no named parameters. Returns: object
Explicitly revive an archived/unpublished entry into the active feed.
Inputs: entry_id (required). Returns: object
Publish an explicit retraction under the same immutable entry ID.
Inputs: entry_id (required), reason. Returns: object
Stop emitting an entry. This does not delete copies already stored by RSS subscribers.
Inputs: entry_id (required). Returns: object
Patch an active entry while preserving its immutable ID. Archived entries require republish_entry first.
Inputs: patch (required), entry_id (required), expected_revision. Returns: object
Strict local RSS/XML validation.
Inputs: no named parameters. Returns: object
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 16 tools
Most tools map cleanly to distinct lifecycle actions: create, update, correct, retract, unpublish, republish, and feed/validation operations are well separated. The main ambiguity is between correct_entry and update_entry, since both modify an entry while preserving its identity; the correction framing helps but could overlap for simple edits.
All tools follow a consistent snake_case verb_noun pattern with clear verbs such as get, list, create, update, validate, audit, and rebuild. Even specialized actions like correct_entry, retract_entry, unpublish_entry, and republish_entry use the same convention, making the set predictable.
Sixteen tools is slightly above the usual sweet spot, but each covers a distinct facet of running an RSS publisher: configuration, entry lifecycle, corrections/retractions, validation/audit, rebuild, batching, and assets. It is not bloated, though a few tools like find_similar_active_entries add extra surface beyond a minimal CRUD set.
The entry lifecycle is well covered with create, read, update, unpublish, republish, correct, retract, and batch publish, plus feed configuration, validation/audit, and rebuild. Minor gaps exist around asset management since only get_asset is exposed and binary upload is handled outside the MCP server; there is also no explicit physical delete, though retract/unpublish appear to be intentional immutable alternatives.
RSS & Atom feed manager with full-text search, AI curation, and remote MCP endpoint.
Managed cloud RSS and Atom feed reader with a remote Model Context Protocol (MCP) endpoint. Zero server setup required — enables AI agents like Claude to instantly subscribe to feeds, search full-text articles, extract readable content, and access curated AI editions ("Your Newspaper") over secure streamable HTTP.
FeedOracle Trust Layer - 10-tool cross-server trust: signing, anchoring, verification.
Verified business OSS MCP for search, RSS, crawling, documents, browser, media and transcription.
