favro-mcp

A Model Context Protocol server for Favro, written in Go.

It speaks MCP over stdio and exposes Favro's REST API as 85 typed tools. Beyond plain CRUD, it ships workflow tools built for natural-language use — search, name→ID resolution, surgical description edits — so an LLM can act on Favro without spending its rate-limit budget on lookup round-trips. Every mutating tool supports dry_run.

Unofficial. Not affiliated with or endorsed by Favro.

Requirements: Go 1.27 to build from source; a Favro account with an API token. See CHANGELOG.md for release history.

Installation

Claude Desktop bundle

Each tagged release publishes a favro-mcp_<version>.mcpb on the GitHub Releases page. Opening it installs the server into Claude Desktop and prompts for the three values it needs — your Favro email, an API token, and the organization id — so there is no config file to hand-edit.

The bundle carries a macOS universal binary, a Windows amd64 binary, and both Linux architectures behind a launcher that picks at startup. It is listed in checksums.txt and covered by the release signature; see Verifying a download.

Cowork plugin (recommended)

Each tagged release publishes a single multi-arch favro-mcp.plugin zip on the GitHub Releases page. The bundle contains binaries for darwin amd64/arm64 and linux amd64/arm64, plus a launcher shim that picks the right one at runtime.

Download favro-mcp.plugin from the release.
Install it into your Cowork profile (drag-and-drop, or follow Cowork's plugin install flow).
Authenticate (see Authentication below).
Restart Cowork — the favro MCP server will appear with all tools registered.

The bundled .mcp.json points at ${CLAUDE_PLUGIN_ROOT}/bin/favro-mcp so the install is relocatable; no host paths to edit.

Standalone binary

For non-Cowork hosts (Claude Code via raw .mcp.json, MCP Inspector, custom integrations):

Download the matching favro-mcp_<version>_<os>_<arch>.tar.gz (or .zip on Windows) from the release.
Extract and place the favro-mcp binary somewhere on $PATH (or note its absolute path).
Authenticate.
Point your MCP host's config at the binary — see MCP host configuration below.

From source

Code

git clone https://github.com/mmedum/favro-mcp
cd favro-mcp
make build       # → bin/favro-mcp

Requires Go 1.27, the version go.mod declares. Under the default GOTOOLCHAIN=auto an older toolchain downloads it on demand; under GOTOOLCHAIN=local the build fails instead of downgrading.

Verifying a download

Every release carries checksums.txt, an SBOM per archive, a keyless cosign signature over the checksum file, and build provenance attested by GitHub. The .mcpb bundle is in checksums.txt with the archives, so one signature covers all of it.

bash

# 1. The file is what the release says it is.
sha256sum -c checksums.txt --ignore-missing

# 2. The checksum file came from this repository's release workflow.
cosign verify-blob checksums.txt \
  --bundle checksums.txt.bundle \
  --certificate-identity-regexp '^https://github\.com/mmedum/favro-mcp/' \
  --certificate-oidc-issuer https://token.actions.githubusercontent.com

# 3. Which workflow, at which commit, built a given artifact.
gh attestation verify favro-mcp_<version>_<os>_<arch>.tar.gz --repo mmedum/favro-mcp

Builds are reproducible: -trimpath plus the commit's own timestamp, so rebuilding a tag gives byte-identical archives.

Authentication

The server uses Favro's HTTP Basic Auth (user email + API token), scoped to one Favro organization at startup. Tools never accept organization_id — pass it once via env var or keyring and forget about it.

API tokens are user-scoped — for team installs, generate the token from a dedicated service-style Favro user with the minimum permissions you need, not from a personal account.

Credential resolution

Two sources, checked in order; the first one that produces a complete (email, token, organization_id) triple wins:

Environment variables — FAVRO_USER_EMAIL, FAVRO_API_TOKEN, FAVRO_ORGANIZATION_ID.
OS keyring — populated once via favro-mcp auth login (cross-platform: macOS Keychain, Windows Credential Manager, Linux Secret Service).

Three more variables change how the server runs rather than who it runs as:

Variable	Effect
`FAVRO_LOG_LEVEL`	`debug` / `info` (default) / `warn` / `error`. Logs go to stderr; stdout carries only the MCP protocol stream.
`FAVRO_MCP_SKIP_VALIDATE`	When set to anything non-empty, skips the startup call that checks the credentials against Favro. For offline testing; the server then fails on the first real tool call instead of at startup.
`FAVRO_ENABLE_DESTRUCTIVE`	Set to `true` to register the delete-style tools. Off by default, and "off" means they are absent from `tools/list` rather than guarded — a host in an auto-approve mode runs a tool without prompting, so not registering it is the only guarantee. Anything the value cannot be read as `true` leaves them off.

`auth` subcommands

Code

favro-mcp auth login       # interactive: masked token input, writes to keyring
favro-mcp auth status      # show user/org (token never printed)
favro-mcp auth logout      # delete keyring entries
favro-mcp auth which       # print active credential source: env or keyring
favro-mcp doctor           # check credentials, org binding and API reachability
favro-mcp doctor --show-ids  # the same, unredacted, for your own screen
favro-mcp --version        # print version + commit
favro-mcp --dry-run        # process-wide override: forces all writes into dry-run

favro-mcp auth login is the one-shot setup for the keyring path. Re-run it to rotate the token.

favro-mcp doctor is the first thing to run when something does not work. It reports the build, which source the credentials came from, whether Favro accepts them, and whether the organization id names one the token can actually see — the failure that otherwise shows up as every tool returning not_found. Its output replaces ids and addresses with stable placeholders so it can be pasted into an issue; --show-ids prints the real values for your own screen, and says so.

MCP host configuration

After the binary or plugin is in place and credentials are stored, register the server in your MCP host config.

Default (keyring path, zero secrets in config):

config.json

{
  "mcpServers": {
    "favro": { "command": "/path/to/favro-mcp" }
  }
}

Headless / CI (env vars) — passes secrets through from the shell, never literal:

config.json

{
  "mcpServers": {
    "favro": {
      "command": "/path/to/favro-mcp",
      "env": {
        "FAVRO_USER_EMAIL": "${FAVRO_USER_EMAIL}",
        "FAVRO_API_TOKEN": "${FAVRO_API_TOKEN}",
        "FAVRO_ORGANIZATION_ID": "${FAVRO_ORGANIZATION_ID}"
      }
    }
  }
}

The Cowork plugin ships its own .mcp.json pointing at ${CLAUDE_PLUGIN_ROOT}/bin/favro-mcp, so for that install path no manual config is needed.

Rate limits

Favro enforces tier-based rate limits per organization (Lite ~100/hr, Standard ~1000/hr, Enterprise ~10000/hr). A misbehaving agent can exhaust the quota for everyone in your org. The server caches resolution lookups (tags, users, widgets, columns) in-memory and supports force_refresh on every list/resolve tool to bust the cache when needed. List tools never auto-aggregate pages — each next_page requires an explicit follow-up call so the LLM can stop early.

Dry-run

Every mutating tool accepts dry_run: true and returns the would-be HTTP request (method, URL, body) plus a state-diff prediction without contacting Favro. Read tools never accept dry_run.

For sandboxed or pre-autonomy testing, --dry-run on the binary forces dry-run mode process-wide regardless of per-call input.

Tools

85 tools covering every Favro REST resource, plus workflow tools built for natural-language use. See docs/TOOLS.md for the full reference.

The ones worth knowing first:

Tool	What it does
`favro_search_cards`	Full-text search over card names and descriptions. Scope to a widget or collection.
`favro_get_card_full`	One card with every id resolved to a name — saves 4–7 follow-up calls.
`favro_resolve_*`	Turn a name into an id for tags, users, collections, widgets, columns, custom fields, groups.
`favro_append_card_description`	Surgical description edits that return a unified diff.
`favro_ping`	Local diagnostic: version, bound org, credential source. Never contacts Favro.

Troubleshooting

Run favro-mcp doctor first — it answers most of the rows below directly, and its output is safe to paste into an issue.

Symptom	Diagnosis & fix
Every tool returns `[not_found]`, but the credentials are accepted	`FAVRO_ORGANIZATION_ID` names an organization this token cannot see. `favro-mcp doctor` reports this as a failed organization binding; `favro-mcp doctor --show-ids` lists the ids the token can see.
`authentication failed — check FAVRO_USER_EMAIL and FAVRO_API_TOKEN env vars` on startup	Either no credentials configured, or the token was revoked / rotated. Run `favro-mcp auth which` to confirm which source the server is reading, then re-run `favro-mcp auth login` (keyring path) or update the env vars.
`FAVRO_ORGANIZATION_ID is required` on startup	The server is single-org by design — it needs to know which org to scope every request to before it can start. Set `FAVRO_ORGANIZATION_ID` (or include it in `auth login`) and restart.
HTTP 429 / `rate limit exceeded`	Hit the per-org Favro rate limit. The client retries once honoring `Retry-After` (capped at 30s) and then surfaces a typed error with `retry_after_seconds`. Use `favro_rate_limit_status` to inspect the most recent `X-RateLimit-*` headers without spending another call.
`next_page` keeps coming back non-null	List tools never auto-aggregate. Each follow-up call must include `page` plus the same `request_id` (Favro routes paginated reads via `X-Favro-Backend-Identifier`); some tools also require resending the original filters.
`decodeJSONLenient` errors with `content-type: text/html, body-prefix: "<p>It looks like…"`	You hit a Favro endpoint that doesn't exist on the documented surface — Favro returns the SPA fallback page instead of a 404. The error includes status + content-type + body-prefix so you can diagnose in one round-trip.
A custom-field write returns 200 but the value doesn't change	Two known causes. First, the field may not be enabled on that widget — org-global custom fields still need per-widget enablement in the Favro UI, and writes to a field the widget doesn't have are accepted and ignored. Second, Favro ignores unknown body fields silently, so a wrong per-type shape looks exactly like success; `favro_set_card_custom_field` sends the shapes Favro's REST docs specify, but they have not all been confirmed against a live tenant. Re-read the card with `favro_get_card_full` to confirm.
Tool descriptions have stale data after a successful write	All cache invalidations are per-org-scoped and per-resource-type; if a write succeeds but the next read still shows old data, pass `force_refresh: true` to the list/resolve tool.
Linux launcher doesn't run from the plugin	The launcher is a bash script. If your shell can't exec it, point your `.mcp.json` directly at `${CLAUDE_PLUGIN_ROOT}/bin/linux-amd64/favro-mcp` (or `linux-arm64`) instead.
Windows	The bundled `bin/favro-mcp.cmd` shim exec's `bin\windows-amd64\favro-mcp.exe`. Windows hosts resolve `${CLAUDE_PLUGIN_ROOT}/bin/favro-mcp` to the `.cmd` automatically via PATHEXT, so the standard `.mcp.json` config in MCP host configuration works as-is. If your host doesn't honor PATHEXT, point `command` at `${CLAUDE_PLUGIN_ROOT}/bin/favro-mcp.cmd` (or directly at `bin/windows-amd64/favro-mcp.exe`).

Documentation


docs/TOOLS.md	Every tool, its inputs and what it returns.
docs/configuration.md	Every setting, the commands, and what happens at startup.
docs/security.md	Trust boundaries, what reaches a log, and what limits what.
docs/development.md	The gates, the test conventions, and how a release is cut.
docs/architecture.md	The design, the decided constraints, the evidence log and the phase plan.
SECURITY.md	Reporting a vulnerability.

Development

See docs/development.md for the gates and the test conventions, and CONTRIBUTING.md for how to get set up and open a pull request.

License

Apache 2.0 — see LICENSE and NOTICE.