v0.0.15

What's Changed

• ci(release): real provenance + SBOM on publish by @sebastienrousseau in #15
• fix(deps): cap mcp below 2.0 to restore FastMCP API by @sebastienrousseau in #16

Full Changelog: v0.0.14...v0.0.15

v0.0.14

Changed

• chore(deps): require bankstatementparser>=0.0.11 (was >=0.0.9),
  keeping the ISO 20022 MCP Suite in lockstep with the 0.0.11 core
  release. No functional or API changes.

v0.0.13

The discoverability cut. Registers bankstatementparser-mcp with
the official Model Context Protocol Registry, adds MCP-spec
conformance CI, and positions the server as part of the ISO 20022 MCP
Suite. No functional or API changes.

Added

• Official MCP Registry integration. bankstatementparser-mcp is
  now registered with the official Model Context Protocol Registry
  (registry.modelcontextprotocol.io) as
  io.github.sebastienrousseau/bankstatementparser-mcp. A new
  server.json at the repo root provides the registry metadata (PyPI
  package identifier, stdio transport), and the README carries an
  mcp-name: io.github.sebastienrousseau/bankstatementparser-mcp
  marker that the registry uses to verify PyPI package ownership.
• Auto-publish workflow (.github/workflows/publish-mcp.yml).
  Authenticates to the MCP Registry via GitHub OIDC (no secrets
  required) on every v*.*.* tag push, syncs the tag version into
  server.json, and runs mcp-publisher publish. Registry metadata
  now stays in lockstep with each PyPI release automatically.
• Protocol conformance CI (.github/workflows/mcp-inspect.yml).
  Runs @modelcontextprotocol/inspector --cli against tools/list
  on every push and PR. Green run is the spec-compliance evidence
  bank buyers routinely ask for.
• Suite discoverability. The README now cross-links the sibling
  banking MCP servers under a "Related MCP Servers" section,
  positioning bankstatementparser-mcp as part of the ISO 20022 MCP
  Suite alongside pain001-mcp, camt053-mcp, acmt001-mcp, and
  noyalib-mcp.

Changed

• GitHub repository description and topics refreshed: description now
  positions the server as part of the ISO 20022 MCP Suite; topics
  extended with mcp, mcp-server, model-context-protocol, iso20022,
  camt053, mt940, ofx, bank-statement, banking, fintech,
  financial-services, claude, ai-agents, python,
  bankstatementparser (topics were empty before).
• Glama manifest version bumped to 0.0.13.

No functional / API changes

• Same 5 MCP tools as v0.0.12. This release is metadata, CI, and
  discoverability only.

v0.0.12

Changed

• Audit pass. Covered main() (removed a # pragma: no cover that masked testable code) and hardened the example-regression discovery to a set-equality check.

v0.0.11

Added

• End-to-end install smoke test - a new smoke job in ci.yml
  ("Smoke test (installed wheel)") builds the wheel, installs it into a
  clean virtual environment (pulling bankstatementparser and mcp
  from PyPI), then imports the installed package and runs an example
  from a neutral working directory so the source tree is not on
  sys.path. This catches packaging and CI-vs-local divergence.
• Expanded edge-case tests keeping the 100% line + branch coverage
  honest: garbage/malformed payloads across the detect/parse/validate
  paths, the limit parameter at its boundaries (0 and larger than
  the row count), an explicit-but-wrong format, and confirmation that
  validate_statement returns a structured {"is_valid": false, ...}
  error rather than raising.

Changed

• Pruned over-scaffolding CI - removed the nightly.yml and
  docs.yml workflows (redundant for a small thin-adapter package);
  the kept workflows (ci.yml, pr.yml, codeql.yml, security.yml,
  release.yml, docker.yml) already cover the full quality posture.

v0.0.10

Added

• Initial release of bankstatementparser-mcp, a Model Context Protocol
  (MCP) server that exposes the
  bankstatementparser
  bank statement parsing library as tools for AI agents and assistants.
• bankstatementparser-mcp console script that runs the FastMCP server
  over stdio.
• Five MCP tools, all delegating to the bankstatementparser parser
  core so they behave identically to the library:
  • list_supported_formats - list every bank statement format the
    parser can read (camt, pain001, csv, ofx, qfx, mt940).
  • detect_format - detect which statement format an inline payload is.
  • parse_statement - parse a statement into structured transactions
    plus a summary, with an optional row limit.
  • validate_statement - dry-run check whether a statement parses
    cleanly, returning a structured result rather than raising.
  • summarize_statement - return only the statement summary
    (opening/closing balances and currency).
• One resource (bankstatementparser://formats) describing each
  supported format and its file extensions, and one prompt
  (analyze_statement) guiding an agent through reading and reconciling
  a statement.
• Tools take inline statement content plus a filename hint and
  materialise it in a private temporary file for the duration of a
  single call, so no shared filesystem is required between client and
  server.
• Multi-stage Dockerfile - a python:3.12-slim-based image runs
  the server over stdio as a non-root mcp user
  (docker run -i --rm bankstatementparser-mcp).
• Quality workflows - ci.yml enforces ruff, mypy, the 100% pytest
  coverage gate, and the 100% docstring gate on Python 3.10/3.11/3.12;
  security.yml runs bandit + pip-audit; codeql.yml runs GitHub's
  CodeQL Python analysis.
• Security policy (SECURITY.md) describing the threat model and
  hardening (private temp files, no network listener, no secrets).
• scripts/verify_versions.py - pre-release script asserting
  __version__, pyproject.toml, and CHANGELOG.md agree.
• Python 3.10+ support; depends on bankstatementparser (>=0.0.9) and
  mcp (>=1.2).
• Quality gates pinned at 100% from the initial release:
  • pytest --cov=bankstatementparser_mcp --cov-branch --cov-fail-under=100
    exercising every line and branch in
    bankstatementparser_mcp/server.py.
  • interrogate --fail-under=100 for module and function docstring
    coverage.
• Runnable examples under examples/ covering every tool, the
  bankstatementparser://formats resource (formats_resource), and the
  analyze_statement prompt — each self-contained (inline fixtures, no
  network) and exercised by the test suite.
• Documentation regression suite keeping the docs honest:
  • tests/test_docs_accuracy.py asserts the README, CHANGELOG, and
    examples/README.md match the live tool/resource/prompt registry,
    the package version, and the supported-format set.
  • tests/test_regression_examples.py runs every examples/*.py as a
    subprocess and asserts a clean exit.
  • tests/test_regression_docs.py executes every fenced python
    block in the README (and docs/*.md) so no documented example can
    silently rot.