Use this server as a unified MCP gateway to discover, validate, generate, and parse ISO 20022 messages across the pain, pacs, camt, and acmt families.
Search the ISO 20022 catalogue by use case, keyword, or message type (e.g. 'reconciliation', 'pacs.008').
List routed families, their capabilities, backing packages, and whether those packages are installed.
List the whole ISO 20022 suite: message families, E&I messages, and specialized servers like reconciliation and agent-payment bridge.
Describe a message type to get its required fields and input JSON Schema.
Validate records against a message type's schema via the family's backing server.
Generate validated ISO 20022 XML from records for pain, pacs, and acmt families, with all missing/invalid fields reported at once on failure.
Parse inbound ISO 20022 XML into structured data for pacs and camt families only.
Route across ISO 20022 families: pain, pacs, camt, acmt, and camt.056/camt.029 exceptions.
Operate over stdio, streamable HTTP, or SSE transports.
Handle missing backing packages gracefully with clear installation instructions.
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., "@iso20022-mcp search for reconciliation messages"
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.
One Model Context Protocol server, one small set of meta-tools —
search, list_families, describe, validate, generate, parse — that
route across every ISO 20022 message family (pain · pacs · camt ·
acmt). Install one thing, discover the whole suite: an agent sees a handful
of verbs instead of the 60+ tools spread across five individual servers.
Latest release: v0.0.13: 7 routing meta-tools over stdio, streamable HTTP or SSE, light core (only
mcp), backing family servers as optional extras, actionable structured errors on everyvalidate/generatefailure, for Python 3.10+. The front door to the ISO 20022 MCP Suite.
The suite has a dedicated, best-in-class server per message family. That depth
is the point — but an agent shouldn't have to know which of five servers and
which of sixty tools it needs before it can act. iso20022-mcp is the thin
routing layer on top: ask it in plain terms ("I need to reconcile a
statement", "generate a credit transfer"), and it points you at — or executes
against — the right family. Small tool surface, whole-suite reach.
Related MCP server: camt053-mcp
iso20022-mcp is the generic message toolkit of eight coordinated,
vendor-neutral MCP servers that together cover the ISO 20022 bank-statement
workflow and the November 2026 structured-address cutover, plus a high-level orchestration layer — readiness scoring, clearing-profile linting, and audit evidence — statement depth,
whole-catalogue routing, reconciliation, multi-format ingestion, and address
remediation. Dependency ranges are kept aligned across the suite,
so the servers co-install cleanly in a single Python environment: start with
one, add the rest as your workflow grows.
Server | Scope | Surface | Install | Use it when |
ISO 20022 | 24 MCP tools · 4 prompts · 3 resources |
| You work with bank-to-customer statements end to end — the suite's flagship | |
Unified gateway: | 7 meta-tools |
| You want one entry point to every message family — this package | |
Matches expected | 7 MCP tools |
| You need explainable statement/payment reconciliation | |
Multi-format statement ingestion: ISO 20022 CAMT.053 and pain.001, SWIFT MT940, OFX/QFX, CSV | 5 MCP tools · 1 prompt · 1 resource |
| Your statements arrive in mixed or legacy formats | |
ISO 20022 postal-address classification, assessment & remediation for the November 2026 structured-address cutover ( | 9 MCP tools |
| You need debtor/creditor addresses cliff-ready ahead of 14 Nov 2026 | |
Orchestration gateway: detect → structurally validate → clearing-profile lint → readiness score, plus automated remediation and | 4 MCP tools |
| You want one high-level readiness / orchestration entry point over the suite | |
Manages, validates and serves bank-specific clearing profiles / rule packs (CBPR+, SEPA_Instant, FedNow, Generic); premium rule-pack entitlement gating | 4 MCP tools |
| You lint payments against your own institution's market practice | |
Compiles readiness findings, remediation diffs and simulated responses into a sealed, Ed25519-signable audit evidence pack | 6 MCP tools |
| You need tamper-evident audit / certification artifacts |
In one line each: camt053-mcp is the bank-statement flagship (deepest
camt.05x surface, stdio + authenticated streamable HTTP);
iso20022-mcp is the generic message toolkit (a handful of verbs over
the whole catalogue); reconcile-mcp is the reconciliation workflow
(did the money we expected actually arrive?);
bankstatementparser-mcp is the ingestion layer (many formats in, one
transaction shape out); and structured-address-fix-mcp is the
postal-address specialist (debtor/creditor addresses cliff-ready for the
Nov 2026 cutover).
The gateway also routes to the per-family servers —
pain001-mcp, pacs008-mcp,
acmt001-mcp, and camt-exceptions —
installed as extras (see Routing).
The core is light. Add the families you need as extras (or [all]):
pip install "iso20022-mcp[all]" # every family
pip install "iso20022-mcp[pacs,camt]" # just interbank + statements
pip install iso20022-mcp # core only; families report as not installed
# or run without installing:
uvx --from "iso20022-mcp[all]" iso20022-mcpMCP client config (e.g. Claude Desktop):
{
"mcpServers": {
"iso20022": {
"command": "iso20022-mcp"
}
}
}One command line, three transports:
Command | Transport | Endpoint | Protocol revisions |
| stdio | the client spawns the process | 2026-07-28, 2025-11-25 |
| Streamable HTTP |
| 2026-07-28 (stateless, |
| HTTP+SSE (2024-11-05) |
| for clients that still expect the older transport |
--host and --port change the bind address (defaults 127.0.0.1 and
8000). The HTTP transports carry no authentication of their own: bind
loopback, or put the server behind a gateway you trust before binding a
routable address. Every release is verified over streamable HTTP with
passmcp in both protocol
eras and over SSE with the MCP SDK client; see
ADR 0001.
{
"mcpServers": {
"iso20022": { "url": "http://127.0.0.1:8000/mcp" }
}
}The 7 routing meta-tools:
search — Find message types by use-case / keyword ("reconciliation", "pacs.008"). (families: all — catalogue)
list_families — List families, their capabilities, and which backing packages are installed. (families: all)
list_servers — Full suite map: families + E&I messages + specialized servers (reconcile, agent-payment bridge). (families: all)
describe — Required fields + input JSON Schema for a message type. (families: all)
validate — Validate records against a message type's schema. (families: all)
generate — Generate a validated ISO 20022 XML message from records. (families: pain · pacs · acmt)
parse — Parse an inbound ISO 20022 XML message into structured data. (families: pacs · camt)
Operations are capability-aware: generate on a camt.053 statement (an
inbound format) returns a clear, explanatory error rather than failing
obscurely, and every "package not installed" case tells you exactly what to
pip install.
Three ergonomics guarantees hold across the routed families:
Actionable structured errors. Any error a backing server raises —
not just the ones it returns — reaches the agent as an
{"error": "<Type>: <message>"} payload instead of an opaque
tool-execution failure. For pain generation the error lists every
missing or invalid field at once, so one fix-and-retry suffices.
Family aliases for message_type. Bare family names resolve to a
concrete version via the backing server: pain.001 →
pain.001.001.09, pain.008 → pain.008.001.02. Fully-versioned
types are still preferred where you know them.
Per-family parse coverage, stated up front. parse covers the
inbound families only: pacs (e.g. pacs.008) and camt (e.g.
camt.053). The initiation families pain and acmt are
outbound-only — they have no parser here, so don't attempt a
generate→parse round-trip for pain.001 or acmt.001; use validate
(or the backing server's XSD validation) instead.
Prefix | Family | Backing server | generate | parse |
| Customer Credit Transfer Initiation | ✅ | — | |
| FI-to-FI Customer Credit Transfer | ✅ | ✅ | |
| Bank-to-Customer Statement | — | ✅ | |
| Account Opening Instruction | ✅ | — | |
| Cancellation / Resolution (E&I) | ✅ | — |
generate("camt.056.001.12", …) routes to camt-exceptions. The gateway also
surfaces the specialized servers — reconcile-mcp and
ap2-iso20022 — via search and list_servers for discovery
(they're invoked through their own tools).
The gateway imports each backing server lazily and optionally — the core
depends only on mcp, and a family's server is loaded on first use. Message
types are matched on their prefix (pacs.008.001.08 → pacs). The pain
extra requires pain001-mcp >= 0.0.56, the release that carries the
LLM-ergonomic generate surface (field aliases, computed nb_of_txs /
ctrl_sum, all-at-once field errors) the gateway's guarantees build on.
search(query="reconciliation")
→ [{ "message_type": "camt.053", "family": "camt", "package": "camt053-mcp" }]
describe(message_type="pacs.008") # required fields + input schema
validate(message_type="pacs.008", records=[…])
generate(message_type="pain.001", records=[…]) # → { "xml": "<Document>…" }
# bare "pain.001" resolves to pain.001.001.09; on failure the payload is
# { "error": "Missing required fields for pain.001.001.09: …" } listing
# every missing or invalid field at once — fix them all and retry once
parse(message_type="camt.053", xml="<Document>…")
# parse covers pacs + camt only; pain and acmt are outbound-onlyFor the dedicated reconciliation engine that matches camt.053 statements
against expected pain.001 payments, see reconcile-mcp.
git clone https://github.com/sebastienrousseau/iso20022-mcp
cd iso20022-mcp
python -m venv .venv && . .venv/bin/activate
pip install -e . && pip install pytest pytest-cov ruff black mypy
pytest # 100% branch coverage gate (backends faked)
ruff check iso20022_mcp tests && black --check iso20022_mcp tests && mypy iso20022_mcpThis project is dual-licensed under either:
Apache License, Version 2.0 (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
MIT License (LICENSE-MIT or https://opensource.org/licenses/MIT)
at your option.
mcp-name: io.github.sebastienrousseau/iso20022-mcp
Describe a message type: its required fields and input JSON Schema, resolved from the family's backing server. When to use: Use after 'search' to obtain the schema and required field names needed to construct flat records before calling 'validate' or 'generate'. When NOT to use: Do not use on raw XML payloads; use 'parse' instead. Behavior: Read-only, idempotent resolution from local schemas. Errors return a structured {'error': ...} payload rather than raising.
| Name | Required | Description | Default |
|---|---|---|---|
| message_type | Yes | ISO 20022 message type, e.g. 'pain.001.001.09', 'pacs.008' or 'camt.053'. For pain messages a fully-versioned type is preferred; the bare family name 'pain.001' resolves to 'pain.001.001.09'. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint, and non-destructive, so the safety profile is covered. The description adds genuinely new context beyond that: resolution from the family's backing server, and that errors return a structured {'error': ...} payload instead of raising. It omits auth/rate-limit details, so 4 rather than 5.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with the core purpose, then labeled sections for usage and behavior. Every sentence carries information, though the 'When to use/When NOT to use/Behavior' scaffolding is slightly formulaic and longer than strictly necessary.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, so return values need no explanation. For a single-parameter read tool, the description covers purpose, prerequisites, alternatives, error semantics, and idempotency — everything an agent needs to invoke it correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% and the single parameter's description already documents the ISO 20022 format and the pain.001 resolution behavior with examples. The description adds no parameter-level detail beyond the schema, so the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (describe) and resource (message type), and names the exact output (required fields and input JSON Schema). It is clearly distinguishable from siblings like 'search', 'validate', 'generate', and 'parse', which are referenced by name.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly structures guidance with 'When to use' (after 'search', before 'validate'/'generate') and 'When NOT to use' (raw XML payloads → use 'parse' instead). Both the positive condition and the alternative are named, leaving nothing to inference.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Generate a validated ISO 20022 XML message from records; the XML document is returned in the 'xml' key. Supported for initiation and interbank families (pain, pacs, acmt); statement families (camt) are inbound-only and return an explanatory error. On failure the 'error' value lists every missing or invalid field at once — fix them all and retry once. When to use: Use to create standardized XML messages for outbound initiation and interbank transfer families (pain.001, pacs.008, acmt.001, camt.056). When NOT to use: Do not use for inbound reporting messages (camt.053); statement families are inbound-only and have no generator. Behavior: Pure transformation; read-only and idempotent with no external side-effects.
| Name | Required | Description | Default |
|---|---|---|---|
| records | Yes | List of flat record objects (field name → value). Field names come from the family's input schema — call describe(message_type) for the required fields. For pain.001 credit transfers the key fields are: id, date ('YYYY-MM-DD' accepted), initiator_name, payment_id, requested_execution_date ('YYYY-MM-DD'), debtor_name, debtor_account_IBAN, debtor_agent_BIC, creditor_name, creditor_account_IBAN, creditor_agent_BIC, payment_amount (alias 'amount'), currency (alias 'payment_currency'), remittance_information; batch_booking takes JSON true/false, and nb_of_txs/ctrl_sum are computed automatically. IBAN and BIC values are strictly validated, never coerced. | |
| message_type | Yes | ISO 20022 message type, e.g. 'pain.001.001.09', 'pacs.008' or 'camt.053'. For pain messages a fully-versioned type is preferred; the bare family name 'pain.001' resolves to 'pain.001.001.09'. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint and destructiveHint=false, so the closing 'read-only and idempotent' line adds little. The real value is the failure contract: errors list every missing/invalid field at once with an instruction to fix all and retry once, plus the inbound-only camt error behavior — neither is derivable from annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with purpose, then scope, error behavior and usage routing. Mildly redundant: the camt inbound-only limitation is stated twice (once in the family sentence, once under 'When NOT to use'), which costs a little without adding information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 2-parameter transformation tool with an output schema, the description covers input routing, failure semantics and family limitations. The mention of the 'xml' return key duplicates the output schema slightly, but nothing an agent needs to invoke it correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%: the records field documents key pain.001 fields, aliases and strict IBAN/BIC validation, and message_type documents versioning and family resolution. The description adds no parameter-level detail beyond what the schema already carries, so the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Generate a validated ISO 20022 XML message from records') plus where the result lands ('the xml key'). It also delineates scope by family (pain/pacs/acmt supported, camt inbound-only), which cleanly separates it from siblings like validate and parse.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit 'When to use' and 'When NOT to use' sections name concrete message types (pain.001, pacs.008, acmt.001, camt.056 supported; camt.053 excluded because statement families have no generator). Nothing about routing is left to inference.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List every ISO 20022 family the gateway routes to (pain, pacs, camt, acmt): its capabilities, backing package, and whether that package is installed in this environment. When to use: Use to check environment readiness and verify whether optional backing packages (pain001-mcp, pacs008-mcp, camt053-mcp, acmt001-mcp) are installed. When NOT to use: Do not use to search specific schemas; use 'search' or 'describe'. Behavior: Pure in-memory introspection; read-only and idempotent.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, idempotentHint=true, destructiveHint=false and openWorldHint=false, so the 'read-only and idempotent' sentence largely restates structured data. The added value is small: 'pure in-memory introspection' clarifies there is no network dependency, and the purpose text implies environment-local results. It does not contradict the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded purpose sentence, then labeled usage and behavior sections — easy to scan and each sentence carries information. The labeled structure is slightly verbose for a simple enumeration tool, but nothing is wasted.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, so return-value detail is unnecessary; the description covers what the tool enumerates, when to reach for it, the alternative tools, and its read-only nature. Nothing an agent needs to invoke it correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Zero parameters, so the schema is trivially complete and the description need not document argument semantics. Baseline 4 applies for a no-parameter tool.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('List every ISO 20022 family the gateway routes to') and enumerates the returned fields (capabilities, backing package, installed status). An agent can distinguish it from list_servers or describe without opening any schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit 'When to use' (environment readiness, verifying optional backing packages) and 'When NOT to use' (schema search) with named alternatives 'search' and 'describe'. Routing is unambiguous.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List the whole ISO 20022 suite the gateway knows: the message families (pain/pacs/camt/acmt), the Exceptions & Investigations messages (camt.056/camt.029), and the specialized servers (reconciliation, agent-payment bridge) with what each does. When to use: Use to discover available specialized servers across the suite and route cross-cutting payments workflows. When NOT to use: Do not use for operational message parsing or generation; call specific gateway tools ('generate', 'parse', 'validate'). Behavior: Pure in-memory catalog listing; read-only and idempotent.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint and destructiveHint=false, and the description largely restates these ('read-only and idempotent'). It does add one genuine piece of context — that it is a pure in-memory catalog with no external fetch — but adds nothing about catalog scope limits, staleness, or rate behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loads the purpose, then usage, then behavior in a clean three-part structure. The enumeration of message families is long but earns its place by defining catalog scope; only the closing 'read-only and idempotent' clause is redundant with annotations.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With an output schema present, return values need not be described, and the description covers purpose, usage and safety. It does not, however, distinguish itself from the sibling 'list_families', which overlaps with the message-family enumeration it advertises.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool takes zero parameters, so the baseline is 4; there is nothing for the description to clarify. It correctly avoids inventing filtering options that the empty schema does not support.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb ('List') and resource (the ISO 20022 suite the gateway knows), and enumerates the concrete contents (pain/pacs/camt/acmt families, E&I messages, specialized servers). An agent can tell this is a discovery/catalog tool distinct from action tools without opening a schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly labeled 'When to use' (discover available specialized servers and route cross-cutting workflows) and 'When NOT to use' (operational parsing/generation), naming the alternative tools 'generate', 'parse', 'validate'. This is the full when/when-not/alternatives pattern.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Parse an inbound ISO 20022 XML message into structured data. Parse coverage is per-family: pacs (e.g. pacs.008) and camt (e.g. camt.053) only. The initiation families pain and acmt are outbound-only — they have NO parser here, so do not attempt a generate→parse round-trip for pain.001 or acmt.001; use 'validate' or the backing server's XSD validation instead. When to use: Use to extract structured transactions, entries, debtor/creditor data, and balances from inbound messages (pacs.008 customer credit transfers, camt.053 statements). When NOT to use: Do not use on outbound initiation families (pain.001, acmt.001). Behavior: Pure in-memory XML parsing; read-only and idempotent.
| Name | Required | Description | Default |
|---|---|---|---|
| xml | Yes | Raw ISO 20022 XML to parse. | |
| message_type | Yes | ISO 20022 message type, e.g. 'pain.001.001.09', 'pacs.008' or 'camt.053'. For pain messages a fully-versioned type is preferred; the bare family name 'pain.001' resolves to 'pain.001.001.09'. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint/idempotentHint/destructiveHint=false, so the 'read-only and idempotent' sentence is largely redundant. However, the description carries real non-annotated behavior: coverage is per-family, and pain/acmt have no parser at all. That asymmetry is the kind of thing structured fields cannot express.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with the core action, then organized into labeled usage/behavior blocks. Slightly repetitive — the family restriction is stated twice (coverage sentence and 'When NOT to use') — but every block is functional and nothing is padding.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, so return structure need not be described. With both required parameters documented at 100% coverage, annotations covering safety, and the family-coverage constraint spelled out, an agent has everything needed to call this correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so baseline is 3. The description goes beyond the schema by limiting which message_type values are actually parseable (pacs/camt only, pain/acmt excluded) and by framing the parameter as family-scoped, which the schema alone does not state.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The first sentence names a specific verb (Parse), resource (inbound ISO 20022 XML message), and output (structured data), then immediately scopes it to pacs/camt families. This lets an agent distinguish it from 'generate' (outbound) and 'validate' without opening either schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit 'When to use' and 'When NOT to use' sections, including concrete message types (pacs.008, camt.053) and a named alternative ('use validate or the backing server's XSD validation'). It also warns against a specific misuse pattern: generate→parse round-trips for pain.001/acmt.001.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search the ISO 20022 catalogue by use-case, message type or keyword (e.g. 'reconciliation', 'make a payment', 'pacs.008') and get the matching message types, their family, and which package provides them. When to use: Use this first to discover which ISO 20022 message type fits your workflow and which backing package supplies it. When NOT to use: Do not use to inspect fields or validate XML; use 'describe' or 'validate' once the message type is known. Behavior: Pure in-memory lookup; read-only and idempotent with zero network or filesystem I/O.
| Name | Required | Description | Default |
|---|---|---|---|
| query | No | Use-case, message type or keyword. Empty = all. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint, openWorldHint=false, destructiveHint=false. The description reinforces with 'pure in-memory lookup; read-only and idempotent with zero network or filesystem I/O,' which adds specificity (no I/O) beyond the boolean hints but largely restates the annotation profile. It also discloses the return shape (matching message types, family, providing package).
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with purpose, then cleanly sectioned into When to use / When NOT to use / Behavior. Slightly verbose in the Behavior sentence, which partially duplicates annotations, but every clause remains useful.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Output schema exists, so return values need not be re-explained, and the description still summarizes them succinctly. With purpose, routing guidance, and behavioral notes combined, an agent has everything needed to invoke 'search' correctly as a discovery step.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100% and the single 'query' param is documented in-schema (including the empty-string default). The description adds the example keyword classes ("reconciliation", "make a payment", "pacs.008") which help the agent phrase queries, but does not extend beyond that. Baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (Search) and resource (ISO 20022 catalogue) and enumerates query dimensions (use-case, message type, keyword) with concrete examples. Distinguishes itself from siblings by naming 'describe' and 'validate' as the tools for post-discovery inspection.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit 'When to use' (first step to discover the fitting message type and backing package) and 'When NOT to use' (field inspection or XML validation - use 'describe' or 'validate' instead). Both the triggering condition and named alternatives are present.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Validate records for a message type against its JSON Schema, via the family's backing server. When to use: Use before 'generate' to verify record field types, required fields, date formats (YYYY-MM-DD), and IBAN/BIC constraints without generating XML. When NOT to use: Do not use to validate raw XML documents; use the backing server's XSD validation tools or 'parse'. Behavior: Pure deterministic validation; read-only and idempotent with zero network I/O.
| Name | Required | Description | Default |
|---|---|---|---|
| records | Yes | List of flat record objects (field name → value). Field names come from the family's input schema — call describe(message_type) for the required fields. For pain.001 credit transfers the key fields are: id, date ('YYYY-MM-DD' accepted), initiator_name, payment_id, requested_execution_date ('YYYY-MM-DD'), debtor_name, debtor_account_IBAN, debtor_agent_BIC, creditor_name, creditor_account_IBAN, creditor_agent_BIC, payment_amount (alias 'amount'), currency (alias 'payment_currency'), remittance_information; batch_booking takes JSON true/false, and nb_of_txs/ctrl_sum are computed automatically. IBAN and BIC values are strictly validated, never coerced. | |
| message_type | Yes | ISO 20022 message type, e.g. 'pain.001.001.09', 'pacs.008' or 'camt.053'. For pain messages a fully-versioned type is preferred; the bare family name 'pain.001' resolves to 'pain.001.001.09'. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint, non-destructive and closed-world. The description adds genuine context beyond them: pure deterministic validation, entirely offline (zero network I/O), and JSON-Schema rather than XSD semantics. It does not describe error/report format, but the output schema likely covers that.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with purpose, then labeled sections (When to use / When NOT to use / Behavior) that make routing scannable. Every sentence carries actionable information with no filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given an output schema exists, the description needn't explain return values, and it covers purpose, routing, validation scope, and offline behavior adequately for an agent to call it correctly on the first try.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so both parameters are fully documented in the schema, which even lists the key pain.001 field names, aliases, and type coercions. The description's mentions of YYYY-MM-DD, IBAN/BIC constraints and message-type resolution largely restate schema content rather than adding new semantics.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (validate) and resource (records for a message type) against an explicit target (its JSON Schema). Clearly distinguishes itself from siblings generate and parse by naming them in the when-not-to-use clause.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit when-to-use (before 'generate' to verify field types, required fields, date formats, IBAN/BIC) and when-not-to-use (raw XML documents; use the backing server's XSD validation tools or 'parse'). The alternative and the selecting condition are both named.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.0.5describegeneratelist_familieslist_serversparsesearchvalidateScored across 7 tools
The two discovery/listing tools (list_servers vs list_families) and the two catalogue tools (search vs describe) have some conceptual overlap, but descriptions clearly delineate them (catalogue-wide routing vs per-family readiness vs discovery vs schema inspection). The action tools validate/generate/parse are cleanly separated by direction and intent.
All names are lowercase consistent verbs, and the two multi-word names (list_servers, list_families) follow a clean verb_noun pattern. Minor deviation between bare verbs (search, describe, validate, generate, parse) and the list_* pair, but still highly predictable.
Seven tools is well-scoped for a gateway/orchestration layer whose job is discovery, schema inspection, and dispatching parse/generate/validate. Each tool earns its place with no filler.
The surface covers the full discovery-to-execution lifecycle (search, describe, list_*, validate, generate, parse). The generate/parse asymmetry across families is intentional and documented, though the absence of any pain/acmt parsing path leaves a slight round-trip gap that agents must work around.
Look up ISO 20022 messages, bank transaction codes, legacy MT replacements, schemes and code sets.
Generate and validate SEPA ISO 20022 XML (pain.001, pain.008) and SWIFT MT103 payment messages.
ISO20022Oracle - 12 ISO 20022 tools: pacs/pain/camt parsing, MX validation, MT migration.
Validate, generate & convert EU e-invoices (UBL, CII, XRechnung, Factur-X) — EN 16931 pre-validated.
