mcpField
R
Recite MCP

Recite MCP

Local
github.com/rivradevFinance0Updated 25w ago

About Recite MCP

Recite MCP is an MCP server connecting AI agents to the Recite API and local receipt bookkeeping. It combines receipt processing and API operations with local ledger and persistent agent memory tools.

Key features

  1. 01Receipt scanning and transaction management
  2. 02Batch scans, projects, and summaries
  3. 03Reconciliation, rules, and exports
  4. 04Local CSV ledger tools
  5. 05Persistent agent memory and health checks

Use cases

  1. 01Process receipts into a local ledger
  2. 02Import or reconcile transaction records
  3. 03Summarize and export ledger data
  4. 04Store persistent agent instructions

MCP support

Exposes Recite API operations and local ledger, memory, configuration, and health tools to MCP clients. Resources include the ledger, memory, and health endpoints.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"rivradev-recite": {
"command": "uvx",
"args": [
"recite-mcp"
],
"env": {
"RECITE_API_KEY": "<RECITE_API_KEY>",
"RECITE_HOME": "<RECITE_HOME>"
}
}
}
}
Run commanduvx recite-mcp
Environment variables to set
RECITE_API_KEYSecretRecite API key used for receipt processing requests.
RECITE_HOMEOptional local data directory for config, ledger, and memory files.

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypirecite-mcp0.1.4

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

recite-mcp

mcp-name: io.github.rivradev/recite-mcp

recite-mcp is an MCP server for the Recite API. It provides the full documented Recite MCP surface for scans, transactions, imports, batch jobs, projects, summaries, webhooks, rules, usage, and exports, while preserving the local ledger/memory workflow for agents that want local-first bookkeeping.

Install

End users

Recommended:

uvx recite-mcp

Alternative (pipx):

pipx install recite-mcp

Alternative (pip):

python -m pip install recite-mcp
Contributors
python -m pip install -e .[dev]
pytest -q
python -m recite_mcp.server

Required Environment

Get an API key at https://recite.rivra.dev/settings/api (includes 30 free scans per month), then set RECITE_API_KEY (required to process receipts; server can still start without it so validate_setup() can report what's missing):

# Windows PowerShell
$env:RECITE_API_KEY="re_live_xxx"
# macOS/Linux
export RECITE_API_KEY="re_live_xxx"

Optional home override:

# Windows PowerShell
$env:RECITE_HOME="C:\path\to\recite-home"
# macOS/Linux
export RECITE_HOME="$HOME/.config/recite"

MCP Client Config

Copy-paste this config:

{
  "mcpServers": {
    "recite": {
      "command": "uvx",
      "args": ["recite-mcp"],
      "env": {
        "RECITE_API_KEY": "re_live_xxx"
      }
    }
  }
}

If installed via pipx or pip, use:

{
  "mcpServers": {
    "recite": {
      "command": "recite-mcp",
      "args": [],
      "env": {
        "RECITE_API_KEY": "re_live_xxx"
      }
    }
  }
}

Troubleshooting

MCP Registry ownership validation failed
  • Symptom: Registry publish returns a PyPI ownership/README validation error.
  • Fix: keep this exact line in packaged README: mcp-name: io.github.rivradev/recite-mcp, then bump version and upload new PyPI release before publishing to Registry.
RECITE_API_KEY missing or invalid
  • Symptom: validate_setup() reports missing_api_key or API requests are rejected.
  • Fix: set RECITE_API_KEY in MCP client env config (preferred) or shell environment.
uvx command not found
  • Symptom: terminal says uvx is not recognized.
  • Fix: install uv first, or use one of the alternatives:
    • pipx install recite-mcp
    • python -m pip install recite-mcp
recite-mcp command not found after install
  • Symptom: command not recognized after pipx/pip install.
  • Fix:
    • For pipx: run pipx ensurepath, then reopen terminal.
    • For pip: run with module entrypoint: python -m recite_mcp.server
MCP client starts but tools are unavailable
  • Symptom: server appears configured but no tools/resources are listed.
  • Fix:
    • Confirm MCP config JSON is valid.
    • Confirm command and args match your install method (uvx vs recite-mcp).
    • Restart MCP client after config changes.
Can't download from PyPI (uvx/pip)
  • Symptom: install/run fails with network/proxy errors.
  • Fix: ensure your environment can reach PyPI (or configure your proxy / custom index). If you're on a locked-down network, you may need to allowlist PyPI or use an internal mirror.
Quick local validation (no MCP client)
  • Run recite-mcp --validate (or uvx recite-mcp --validate) to print local config/health JSON. Exit code is 0 if an API key is present, otherwise 1.

Main Features

API tools (41 total):

  • Scans — scan_receipt (file, URL, base64, or raw text; ephemeral mode), get_scan
  • Transactions — create, list (with filters), get, update, delete, import_transactions (JSON list, CSV text, or CSV file)
  • Batch scans — submit_batch_scans (1–20 items, async, webhook notification), get_batch_scan_status, get_batch_scan_results
  • Projects — list, create, update, delete
  • Summaries — get_summary (period, date range, group-by)
  • Webhooks — create, list, delete (transaction.created/updated/deleted, batch.completed)
  • Reconciliation — create_reconciliation_link, list_reconciliation_links, update_reconciliation_link, delete_reconciliation_link, run_auto_match, get_reconciliation_summary, get_reconciliation_recommendations, export_reconciliation
  • Rules — create, list, update, delete (vendor_category, default_project, processing_preference)
  • Categories — get_categories (17 built-in + custom), create_category, delete_category
  • Vendors — get_vendors, create_vendor, delete_vendor
  • Usage — get_usage (period, breakdown)
  • Exports — export_transactions (CSV/JSON, optional file save)

Local tools:

  • process_receipt — scan + ledger append in one step (optional rename, dry-run, category hint)
  • process_receipts_batch — scan a directory of images/PDFs (recursive, dry-run preview)
  • summarize_ledger, export_ledger, add_ledger_correction — local CSV ledger management
  • update_memory, list_memory — persistent agent instruction storage
  • get_config, validate_setup — configuration inspection and health check

Resources: recite://ledger, recite://memory, recite://health

Documentation

  • User guide: docs/user_guide.md

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.