Watchlist tracking
Reviewed for relevance and future tracking value.
MCP100 Assessment
Watchlist / Tracking
Why we’re tracking it
Track agent governance and tool-control changes as Trinity evolves
Reviewed · Documentation review; complete setup workflow not independently tested
Best fit
Self-hosting and operating fleets of isolated agents from one control plane.
What we know
Sovereign agent platform with Docker-isolated agents, web/API administration and an MCP control plane for deployment, messaging, schedules and logs.
Maintenance / activity evidence
- Observed (UTC)chore(deps-dev): bump brace-expansion in /tests/git sync ( 3198)
Tracking evidence observed Oct 5, 2026. These dated signals do not establish current maintenance. Reviewed enrichment confidence: high.
Verification gaps / watch-outs
- Installation and comprehensive security have not been independently audited.
- Agent-control-plane authority. Writes: Create/deploy/start/stop/delete agents; Schedules; Messages; GitHub sync.
- Admin/MCP API key + model credentials.
- Sensitive access: Provider credentials, GitHub PATs, Agent workspaces/logs
- Control-plane operations are inherently high authority despite container isolation and audit features.
Full authority, access and evidence limits are available in Review context and Access & risk.
Popularity
583 repository stars
Stars observed Sep 25, 2026.
Do not overwrite existing reviewed stars/download metrics with nulls from this enrichment batch.
Do not overwrite existing reviewed stars/download metrics with nulls from this enrichment batch.
All limitations / Review context
Which MCP integrations and governance controls does the current platform document?
The bounded assessment reviewed the official README, repository notices, security policy path, and security page for known material public-tracking blockers. No such blocker was identified. This is a public-tracking assessment, not an installation safety certification.
- Control-plane operations are inherently high authority despite container isolation and audit features.
- Agent-control-plane authority.
- Admin/MCP API key + model credentials.
- Self-hosted agent/session/log data plus credentials.
- Reviewed write actions: Create/deploy/start/stop/delete agents; Schedules; Messages; GitHub sync
- Sensitive access: Provider credentials, GitHub PATs, Agent workspaces/logs
- Installation, client compatibility and comprehensive security have not been independently audited. Authentication and access depend on the configured services.
- No independent runtime, installation, or comprehensive security audit was performed.
- No installation, runtime, client matrix, or complete tool-contract audit was performed in this review.
- Requires Docker/Compose and provider credentials.
- MCP API key provisioning and admin access are privileged.
- SQLite is a local default; PostgreSQL is recommended for production.
Original activity observations
- 2026-10-03T09:14:55Z: chore(deps-dev): bump brace-expansion in /tests/git sync ( 3198)
Technical identity
github.com/abilityai/trinity- Exact implementation source
- https://github.com/abilityai/trinity ↗
- Reviewed resource type
- Host
- Repository identity
- abilityai/trinity
- Service endpoint
- http://localhost:8080/mcp
Evidence & Sources
- Evidence basis
- Recorded source evidence
- Setup runtime verification
- Not independently runtime-tested by MCP100 for this setup workflow.
Reviewed · Latest source observation
- GitHub repository ↗
abilityai/trinity
Recorded repository material and project history.
Includes references whose attribution is not established.
Setup · Watchlist
Self-hosted gateway service setup
Publisher instructions.
Source evidence observed .
Additional setup details required — see the specific gaps below.
Download / install
Publisher instructionsSelf-host Trinity first; then connect to its MCP HTTP endpoint using an API key created in Trinity Settings. Do not confuse the trinity-cli package with the server stack.
Advanced configuration · self hosted install or run
self hosted install or run · Review before use
git clone https://github.com/abilityai/trinity.git cd trinity cp .env.example .envAdvanced configuration · self hosted start
self hosted start · Review before use
./scripts/deploy/build-base-image.sh ./scripts/deploy/start.shConfigure application / integrations
Publisher instructionsSet ADMIN_PASSWORD (publisher requires 12+ characters) in .env and satisfy Docker/runtime prerequisites. First-run web UI at http://localhost. In Settings → MCP Keys create an MCP API key. Configure HTTP MCP endpoint http://localhost:8080/mcp for local installations, or https://your-domain.com/mcp behind the production proxy. Publisher Claude Code JSON uses mcpServers.trinity.type=http and an Authorization Bearer header. Never put a real API key in committed repository config.
Advanced configuration · claude code json
{ "mcpServers": { "trinity": { "headers": { "Authorization": "Bearer REPLACE_WITH_TRINITY_MCP_KEY" }, "type": "http", "url": "http://localhost:8080/mcp" } } }Check connection
Suggested checkCheck the deployed Trinity control UI Settings > MCP Keys, the appropriate /mcp URL, authentication scope and client-discovered tools. A configured JSON alone does not demonstrate remote key validity. Do not run create/stop/delete agent operations for smoke testing. Client registration check: `claude mcp get trinity` (checks registration; inspect tools/read-only calls separately to verify authenticated service access).
Disconnect / uninstall / revoke
MCP100 setup guideFor a local Trinity installation run `./scripts/deploy/stop.sh` from the checkout to stop services without removing agent-network containers. The publisher explicitly warns against routine `docker compose down`. Separately disconnect client config and revoke Trinity MCP key via Settings → MCP Keys if access should end; do not assume stopping stack revokes keys or removes persistent data. Client disconnect (claude-code, only if registered using the name `trinity`): `claude mcp remove trinity`. This unregisters the client entry; it does not uninstall packages or revoke remote service credentials.
To revoke the Trinity MCP key, use Trinity Settings → MCP Keys to deactivate/revoke the key used for this connection. Revocation is separate from stopping the stack or deleting a client entry.
claude mcp remove trinity
Official Documentation & Sources
- https://github.com/abilityai/trinity/blob/main/README.md ↗
- https://docs.ability.ai/guides/mcp-integration ↗
- https://docs.ability.ai/getting-started/setup ↗
- https://github.com/Abilityai/trinity/blob/main/scripts/deploy/stop.sh ↗
- https://github.com/Abilityai/trinity/blob/main/docs/user-docs/guides/deploying-trinity.md ↗
- https://github.com/abilityai/trinity ↗
What it does
Sovereign agent platform with Docker-isolated agents, web/API administration and an MCP control plane for deployment, messaging, schedules and logs.
- Capability
- AI & Reasoning
- Resource type
- Host
- Best fit
- Self-hosting and operating fleets of isolated agents from one control plane.
Representative uses
- Create/deploy/start/stop/delete agents
- Chat with agents and inspect history/logs
- Manage GitHub sync/PATs
- Deploy multi-agent systems
Limitations
- Requires Docker/Compose and provider credentials.
- MCP API key provisioning and admin access are privileged.
- SQLite is a local default; PostgreSQL is recommended for production.
Reviewed access
Access & risk
Review the exact access boundary before connecting this MCP to accounts, credentials, or sensitive data.
- Authentication
- Admin/MCP API key + model credentials.
- Permissions
- Agent-control-plane authority.
- Data handling
- Self-hosted agent/session/log data plus credentials.
- Cost
- free
- Risk and cautions
- Control-plane operations are inherently high authority despite container isolation and audit features.
Access scope and cautions
- Write actions
- Create/deploy/start/stop/delete agents Schedules Messages GitHub sync
- Sensitive access
- Provider credentials GitHub PATs Agent workspaces/logs
- Risk level
- high
FAQ
Based on reviewed evidence as of 2026-10-07. Watchlist tracking is not a recommendation or a security certification.
Which tasks are documented for this implementation?
What setup and authentication are documented?
Can it modify data or perform write operations?
What data-handling boundary should I consider?
What limitations and risks should I check before use?
Is this an official MCP?
Ability AI is listed as First-party MCP in MCP100’s current evidence.