Arbiter MCP server
Commit Nov 11, 2026Provides a capability firewall that enables AI agents to interact with services like SMTP and IMAP through policy-controlled access.
Also known as omry/arbiter
- Stars
- 3
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 11, 2026
- License
- MIT
What is Arbiter MCP?
Arbiter provides a capability firewall that enables AI agents to interact with services like SMTP and IMAP through policy-controlled access.
Summary generated from the project's README
About Arbiter MCP server
Securing AI agent interactions with email services is what Arbiter enables through a capability firewall. It provides a controlled way for agents to access SMTP and IMAP services using explicit account policies. This prevents unrestricted access by ensuring every action is governed by predefined rules and metadata.
Developers use the Arbiter server via a native HTTP API and manage it using the arbiter-server CLI. The system supports SMTP submission with configured sender identities and TLS settings, alongside IMAP tools for managing folders and messages. Users can interact with the running service through the native arbiter client CLI or by pointing local integrations like VS Code to the server URL.
Setup involves bootstrapping configurations and managing service accounts through specialized objects like arbiter.account and arbiter.policy. The architecture supports Docker deployments and allows for granular control over specific operations such as searching, moving, or deleting IMAP messages within scoped folders.
What you can do with it
- 1Controlled email transmissionAn AI agent can send emails with specific sender identities and HTML bodies through a secure SMTP interface.
- 2Managed inbox interactionA developer can allow an agent to search, read, and organize emails within specific IMAP folders.
- 3Policy-driven service accessAdministrators can define strict access rules for different service accounts using specialized policy objects.
Written from the project's README. Every use case is checked against a quote from it.
Configuration
IMAP_PRIMARY_ACCOUNT_HOSTIMAP_PRIMARY_ACCOUNT_PASSWORDIMAP_PRIMARY_ACCOUNT_USERNAMESMTP_PRIMARY_ACCOUNT_HOSTSMTP_PRIMARY_ACCOUNT_PASSWORDSMTP_PRIMARY_ACCOUNT_USERNAME
| Name | Description | |
|---|---|---|
| IMAP_PRIMARY_ACCOUNT_HOST | — | |
| IMAP_PRIMARY_ACCOUNT_PASSWORD | — | |
| IMAP_PRIMARY_ACCOUNT_USERNAME | — | |
| SMTP_PRIMARY_ACCOUNT_HOST | — | |
| SMTP_PRIMARY_ACCOUNT_PASSWORD | — | |
| SMTP_PRIMARY_ACCOUNT_USERNAME | — |
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableomry/arbiter is public and reachablegithub.com/omry/arbiter ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-11github.com/omry/arbiter ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubArbiter
| Description | |
|---|---|
| Meta packages | |
| Core packages | |
| Plugins | |
| Python | |
| Docs, support, and ecosystem | |
| General |
Arbiter is a capability firewall between AI agents and services. Today it exposes policy-controlled access through an Arbiter server and client CLI; additional interfaces may be added later. The current service surface covers sending mail over SMTP and reading IMAP folders through explicit account policies.
Project Status
Current implementation status:
- Arbiter server over a native HTTP API
- capability discovery with SMTP and IMAP account and operation metadata
- SMTP submission with configured sender identity, TLS/auth settings, text/HTML bodies, and Bcc kept out of message headers
- IMAP list/get/search/move/mark-read/delete tools scoped to configured accounts and folders
arbiter.account.<service>and reusablearbiter.policy.<service>objects for per-service account policy- native
arbiterclient CLI andarbiter-serverserver/operator CLI - Reploy blueprint material for staged and installed Docker deployments
Known open gaps:
- durable audit storage is parked for post-v1, while startup/runtime logging is the v1 observability focus
- normalized error-code responses are still a design contract, while the implementation currently surfaces Python and transport errors
- the agent-facing skill integration path is intentionally not implemented in this repository yet
Development
Create and use the repo-local virtualenv with:
python3 -m venv .venv.venv/bin/python -m pip install --upgrade pip.venv/bin/python -m pip install -r requirements-dev.txt
The repository root is a workspace, not an Arbiter runtime package. The dev
requirements file installs server, plugins/smtp, and plugins/imap
editably so the arbiter-server command and service plugins come from this
checkout.
Install the native client separately when you need the repo-local arbiter
command:
.venv/bin/python -m pip install -e client
Run the test suite from the repo root with:
.venv/bin/python -m nox -s tests.venv/bin/python -m nox -s lint
The lint session runs both black --check and pyrefly check.
For focused local runs without nox, use the same environment directly, for example:
.venv/bin/python -m pytest server/tests/unit/test_config.py.venv/bin/python -m pytest server/tests/unit/test_app.py
GitHub Actions references
External actions use full commit SHAs with version comments. Reusable workflows
also use full SHAs, with one explicit exception: the two Node maintenance calls
in .github/workflows/dependency-audit.yml follow
hydra-ecosystem/github-actions@v1 for automatic compatible fixes. This trusts
the release tag maintained by that repository. A different source, workflow,
caller, branch, or major release requires a separate policy update. Full SHA
references remain valid for maintenance too. See the shared
release policy.
The Docusaurus website lives in website/:
cd website && npm installcd website && npm run startcd website && npm run build
The native client implementation lives in client/go-cli/. Build all default release targets from the repo root with:
tools/build_go_client --clean
This writes stripped Linux, macOS, and Windows binaries for amd64 and arm64
under client/go-cli/dist/. Limit the matrix with one or more --target GOOS-GOARCH arguments, for example tools/build_go_client --target linux-arm64; pass --debug to keep debug symbols.
Package the platform-neutral agent skill with:
tools/build_release_dists --packages skill
This writes the platform-neutral arbiter-skill wheel under dist/. The skill
declares arbiter-client as an Agent Skill Installer companion wheel; ASI lets
pip select the current platform's native client wheel during install.
The native client can also be packaged as the arbiter-client PyPI project:
tools/build_release_dists --packages client
This builds one platform-tagged wheel per native target. Each wheel installs the
native arbiter executable directly as a wheel script; it does not contain a
Python wrapper.
The user-facing documentation lives in website/docs/. The root docs/ directory contains internal planning, security assessments, active designs and an archive of superseded drafts:
- docs/BACKLOG.md
- docs/testing_backlog.md
- docs/release-readiness.md
- Security assessment
- docs/future/
- Historical designs
Local Native HTTP Run
For local Codex or VS Code integration, run Arbiter over native HTTPS and point the client at:
https://127.0.0.1:8075
Arbiter does not ship a runnable service config. Bootstrap a Hydra
config, edit it, then run the server. Keep server-owned files together in a
deployment directory; for local development, arbiter-server/conf is a useful
repository-local config directory.
Plugin-owned object templates are created by the plugin command surface:
arbiter-server --config-dir ./arbiter-server/conf bootstrap --server
arbiter-server --config-dir ./arbiter-server/conf bootstrap --plugin smtp --account primary
${oc.env:...} reads the process environment that your shell, supervisor,
container runtime, or secret manager provides. For local runs, the root config
can name one optional dotenv-style file to load before composition.
See website/docs/operate/configuration-model.md for the generated file layout and composition flow.
For local development, the server env file lives beside conf/:
# arbiter-server/.arbiter.env
SMTP_PRIMARY_ACCOUNT_HOST=smtp.example.com
# SMTP_PRIMARY_ACCOUNT_PORT=587
SMTP_PRIMARY_ACCOUNT_USERNAME=agent@example.com
SMTP_PRIMARY_ACCOUNT_PASSWORD=change-me
IMAP_PRIMARY_ACCOUNT_HOST=imap.example.com
# IMAP_PRIMARY_ACCOUNT_PORT=993
IMAP_PRIMARY_ACCOUNT_USERNAME=agent@example.com
IMAP_PRIMARY_ACCOUNT_PASSWORD=change-me
Point the root config at the env file:
arbiter:
env_file: ../.arbiter.env
Existing process environment variables take precedence over values from the env file. Relative paths are resolved from the config directory.
Build or refresh that file from the active config:
arbiter-server --config-dir ./arbiter-server/conf env bootstrap
arbiter-server --config-dir ./arbiter-server/conf env check
env bootstrap keeps existing assignments, adds missing config references, and
groups entries under sorted # arbiter-* blocks plus # miscellaneous.
If arbiter.env_file is not set yet, it adds arbiter.env_file: ../.arbiter.env to the root config first.
Then run from this directory:
arbiter-server --config-dir ./arbiter-server/conf config check
arbiter-server --config-dir ./arbiter-server/conf serve
config check and serve require at least one configured service account.
Use arbiter-server plugins list to inspect installed service plugins before
validating a config. Once the server is running, point the client CLI at the
server URL. The client accepts Arbiter's local self-signed TLS certificates by
default; configure arbiter.tls_ca_file when you want certificate verification
against a specific certificate authority file:
arbiter info server arbiter.url=https://127.0.0.1:8075
arbiter plugins arbiter.url=https://127.0.0.1:8075
The client can also read the endpoint from a small config file:
~/.arbiter/arbiter-client.yaml.
arbiter:
url: https://127.0.0.1:8075
Override config values with Hydra-style key=value arguments after the
command, or bootstrap the client config:
arbiter bootstrap client arbiter.url=https://127.0.0.1:8075
IMAP operations use folder-scoped UIDs returned by imap:list_messages and
imap:search_messages; pass those ids back to imap:get_message,
imap:move_message, imap:mark_message_read, or imap:delete_message with
the same account and folder.
License
Arbiter is distributed under the MIT License. See LICENSE.
Arbiter MCP questions
What services does Arbiter currently support?
The current service surface includes sending mail over SMTP and reading IMAP folders through explicit account policies.
Does Arbiter provide audit storage?
Durable audit storage is currently parked for post-v1, but startup and runtime logging are the focus for v1 observability.