How do I use fastmcp-fleet?

1. Click on "Deploy Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@fastmcp-fleet simulate a bug triage: search GitHub issues, create a Jira ticket, notify Slack" 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.

fastmcp-fleet

by deepakdalal1221

Overview Schema Source Related Servers Discussions Admin

Python

Hybrid

Provides tools for managing secrets, passwords, and vaults through 1Password.

Allows interaction with Airtable bases, tables, and records.

Enables creating and listing tasks in Asana projects.

Provides authentication and user management capabilities through Auth0.

Provides search functionality via Brave Search.

Allows triggering and monitoring CircleCI pipelines.

Provides tools for managing Cloudflare zones, DNS, and other services.

Allows interaction with Confluence pages and content.

Provides monitoring and observability tools through Datadog.

Allows managing DigitalOcean droplets, resources, and more.

Provides tools for sending messages and interacting with Discord channels.

Allows querying and indexing data in Elasticsearch.

Provides tools for managing Gitea repositories and issues.

Enables creating and listing issues, managing repositories, and other GitHub operations.

Provides tools for managing GitLab repositories, merge requests, and more.

Allows interaction with Grafana dashboards and data sources.

Provides tools for managing contacts, deals, and other HubSpot CRM data.

Allows triggering and managing Jenkins jobs and builds.

Enables creating and searching Jira issues, managing projects.

Provides tools for managing Linear issues and projects.

Allows managing MailChimp campaigns, lists, and subscribers.

Provides tools for interacting with MongoDB databases and collections.

Allows querying and managing MySQL databases.

Provides tools for deploying and managing Netlify sites.

Enables creating pages and searching Notion workspaces.

Provides user management and authentication capabilities through Okta.

Provides mock integration with OpenAI's API for AI completions and models.

Allows creating and listing incidents in PagerDuty.

Provides monitoring and alerting capabilities via Prometheus.

Allows interacting with Redis data structures.

Provides tools for deploying and managing services on Render.

Allows sending emails and managing SendGrid templates.

Provides error tracking and monitoring through Sentry.

Enables posting messages and retrieving conversations in Slack.

Provides security vulnerability scanning and monitoring through Snyk.

Allows querying and managing SQLite databases.

Enables creating payment intents and listing charges in Stripe.

Provides infrastructure management capabilities through Terraform.

Allows sending SMS and listing messages via Twilio.

Provides secrets management and access control through HashiCorp Vault.

Provides tools for deploying and managing Vercel projects.

fastmcp-fleet

License: MIT Python 3.12+ Tests Offline First Servers FastMCP Docker PRs Welcome CodeQL

A local-first FastMCP platform of 160+ mock MCP servers — GitHub, Slack, Stripe, Jira, Notion, Cloudflare, Datadog, and more — running behind one central gateway. Zero live API calls. Deterministic. Perfect for agent trajectory testing and OpenHands integration.

Related MCP server: scv-triage

What is fastmcp-fleet?

fastmcp-fleet is a catalogue of 175 MCP servers — 170 active + 5 pilot, packaged behind one FastMCP gateway on port :8000. Every server can run fully offline — HTTP requests are intercepted by a custom transport that serves fixture JSON, and stateful tools persist to a per-server SQLite store so subsequent reads see prior writes.

The point is not to replace live SaaS APIs. It is to give AI agents (OpenHands, Claude, Cursor, custom stacks) a realistic, deterministic, credential-free environment for:

MCP tool discovery testing
Agent trajectory generation and validation
Multi-step workflow rehearsal (create issue → update task → post message)
CI/CD integration testing without hitting external quotas
Local development and demonstration

Why offline-first?

Real APIs	fastmcp-fleet
Requires credentials for 40+ SaaS accounts	No credentials required
Rate-limited by external providers	No rate limits
Costs money (OpenAI, Anthropic, Stripe, ...)	Free
Non-deterministic (state changes between test runs)	Deterministic (fixture + SQLite)
Slow (network latency)	Fast (µs response times)
Needs internet	Runs in an air-gapped container

Set MCP_OFFLINE=1 (already the default in docker-compose.yml) and every server becomes a convincing local simulation.

Architecture

graph TD
    Agent[OpenHands / Claude / Custom Agent]
    Gateway[FastMCP Gateway<br/>:8000<br/>Bearer auth + rate limit]
    
    Agent -->|MCP / streamable-http| Gateway
    
    Gateway --> GH[github :8001]
    Gateway --> JR[jira :8071]
    Gateway --> SL[slack :8061]
    Gateway --> ST[stripe :8154]
    Gateway --> NT[notion :8081]
    Gateway --> More[... 44 more]
    
    GH -->|HTTP intercept| OT[OfflineTransport]
    OT --> FX[fixtures/*.json]
    OT --> LS[state.db — SQLite]
    
    style Gateway fill:#4a5568,stroke:#2d3748,color:#fff
    style OT fill:#ed8936,stroke:#c05621,color:#fff
    style FX fill:#38a169,stroke:#276749,color:#fff
    style LS fill:#38a169,stroke:#276749,color:#fff

Each MCP server runs in its own container on its own port (8001–8173). The gateway aggregates all of them so an agent connects once and gets the full fleet.

Quickstart

Docker (one container, one command — recommended)

The whole fleet (175 servers, 539 tools, ~200 seeded mock rows) as one image on one port.

git clone https://github.com/deepakdalal1221/fastmcp-fleet.git
cd fastmcp-fleet
make start

make start builds docker/Dockerfile.mono (first time only, ~5 min) and runs it on http://localhost:9000/mcp. Point Claude Desktop / Cursor / any MCP client at that URL and every server is reachable via namespaced tools (github_list_issues, stripe_list_charges, etc.).

make logs-mono     # tail startup logs (wait for 'all_in_one.ready mounted=175')
make stop          # shut down
make restart       # rebuild & restart

State persists in ./state/ across restarts; delete that dir to reset all mock data.

Local (Python + uv)

git clone https://github.com/deepakdalal1221/fastmcp-fleet.git
cd fastmcp-fleet
uv sync --extra all --extra dev
uv run python -m pytest tests/ -q
# 200 passed in ~1s

Run one server directly:

export MCP_OFFLINE=1
uv run python -m servers.github.server --port 8001

Then point Claude Desktop or any MCP client at http://localhost:8001/mcp.

Docker (all 55 services + gateway)

cd deployment/compose
docker compose up -d --build
# Gateway available at http://localhost:8000

The compose stack ships with MCP_OFFLINE: "1" set on every service via the x-common-env anchor, and a ./state:/state volume so SQLite state persists across restarts.

Server catalogue

170 active + 5 pilot = 175 servers across 13 categories. Full list of 175 catalogued in registry/servers/.

Category	Count	Servers
AI / ML	5	`anthropic`, `gemini`, `huggingface`, `mistral`, `openai`
Business	3	`hubspot`, `mailchimp`, `stripe`
Cloud	6	`cloudflare`, `digitalocean`, `fly-io`, `netlify`, `render`, `vercel`
Communication	4	`discord`, `sendgrid`, `slack`, `twilio`
Database	5	`elasticsearch`, `mongodb`, `mysql`, `redis`, `sqlite`
Development	2	`gitea`, `gitlab`
DevOps	4	`circleci`, `gitlab-ci`, `jenkins`, `terraform`
Monitoring	6	`datadog`, `grafana`, `new-relic`, `pagerduty`, `prometheus`, `sentry`
Productivity	3	`airtable`, `confluence`, `notion`
Project Management	3	`asana`, `jira`, `linear`
Search	2	`brave-search`, `tavily`
Security	5	`1password`, `auth0`, `okta`, `snyk`, `vault`
Storage	1	`s3`

Plus foundational: fetch, filesystem, time, github, postgres.

How offline mode works

sequenceDiagram
    participant Agent
    participant Server as MCP server<br/>(e.g. github)
    participant OT as OfflineTransport
    participant FS as fixtures/*.json
    participant DB as state.db (SQLite)
    
    Agent->>Server: tool_call(list_issues, owner=x, repo=y)
    Server->>Server: is_offline() → True
    Server->>DB: local_store.list_all("github", "issues:x/y")
    DB-->>Server: [issue1, issue2]
    Server-->>Agent: {"issues": [issue1, issue2]}
    
    Note over Agent,DB: Non-mutating tools fall through to OfflineTransport
    
    Agent->>Server: tool_call(get_repo, owner=x, repo=y)
    Server->>OT: httpx GET /repos/x/y
    OT->>FS: read fixtures/get_repos_x_y.json
    FS-->>OT: JSON payload
    OT-->>Server: 200 OK + payload
    Server-->>Agent: {"id": ..., "full_name": ...}

Layered lookup: specific fixture → default.json → synthetic {"offline": true, ...}. Add per-endpoint JSON files at servers/<id>/fixtures/<method>_<path-slug>.json to customize responses.

Local Store — stateful mocks

Nine servers wire mutation tools to a SQLite-backed writable state so agent workflows produce consistent results:

stateDiagram-v2
    [*] --> Empty
    Empty --> HasIssue: github.create_issue()
    HasIssue --> HasIssue: github.list_issues() → returns created
    HasIssue --> HasMore: jira.create_issue()
    HasMore --> HasMore: slack.post_message()
    HasMore --> Verified: chain_test asserts all reads

Stateful servers today:

Server	Write tool	Read tool	State key
github	`create_issue`	`list_issues`	`issues:<owner>/<repo>`
jira	`create_issue`	`search_issues` (JQL)	`issues:<project_key>`
asana	`create_task`	`list_tasks`	`tasks:<project_id>`
linear	`create_issue`	`list_issues`	`issues:<team_id>`
slack	`post_message`	`get_conversation`	`messages:<channel>`
stripe	`create_payment_intent`	`list_charges`	`charges`
twilio	`send_sms`	`list_messages`	`messages:<to>`
pagerduty	`create_incident`	`list_incidents`	`incidents:<service_id>`
notion	`create_page`	`search`	`pages:<parent_id>`

State survives docker compose down/up via the ./state:/state volume mount.

Add a new server in 5 minutes

# 1. Add an Entry to CATALOGUE (append at end — port assignment is index-based)
#    Edit scripts/build_registry.py

# 2. Regenerate manifests
uv run python scripts/build_registry.py

# 3. Scaffold the server
uv run mcp-create <your-server-id>

# 4. Implement servers/<your-server-id>/tools.py following the pattern in
#    servers/github/tools.py or servers/cloudflare/tools.py

# 5. Promote status: planned → active in registry/servers/<id>.yaml

See CONTRIBUTING.md for the full recipe (auth patterns, error mapping, response slimming, Local Store wiring).

Testing

uv run python -m pytest tests/ -q

........................................................................ [ 36%]
........................................................................ [ 72%]
........................................................                 [100%]
200 passed in 1.10s

Breakdown:

190 contract tests validate every manifest (schema, unique ports, valid auth types) and every server file (importable, register_tools exists, tools registered)
10 offline trajectory tests exercise create → read chains across the 9 stateful servers, plus one cross-server chain (github issue → jira issue → slack message)

Docker deployment

55 services total (gateway + 54 servers). Each server:

Built from docker/Dockerfile.server with SERVER_ID build-arg
Exposes its port (8001–8173)
Inherits MCP_OFFLINE: "1" and MCP_STATE_DIR: /state from x-common-env
Mounts ./state:/state for persistence (47 of them; gateway and filesystem opt out correctly)
Registered in the gateway via GATEWAY_SERVERS env

graph LR
    subgraph "docker-compose stack"
        GW[gateway :8000]
        subgraph "network: mcp"
            GH[github :8001]
            GL[gitlab :8002]
            SL[slack :8061]
            ST[stripe :8154]
            OA[openai :8134]
            More[... 44 more]
        end
        VOL[./state:/state<br/>persistent volume]
    end
    GW -.mounts.-> GH & GL & SL & ST & OA & More
    GH & GL & SL & ST & OA & More -.state.-> VOL
    
    style GW fill:#4a5568,color:#fff
    style VOL fill:#38a169,color:#fff

CRUD coverage

Per-server capabilities. Create · Read · Update · Delete. · = not implemented.

Server	Category	C R U D	Tools
`1password`	security	`· R · ·`	list_vaults, list_items, get_item
`airtable`	productivity	`C R U D`	create_record, delete_record, list_bases, list_records, list_tables, update_reco
`ansible`	devops	`· R · ·`	list_inventories, list_job_templates, list_projects
`anthropic`	ai-ml	`· R · ·`	messages
`argocd`	devops	`· R U ·`	get_application, list_applications, sync_application
`asana`	project-management	`C R · ·`	list_workspaces, list_projects, list_tasks, create_task
`auth0`	security	`· R · ·`	list_users, list_clients, list_rules
`azure-devops`	development	`· R · ·`	list_pipelines, list_projects, list_repos
`bitbucket`	development	`· R · ·`	get_repository, list_pullrequests, list_repositories
`brave-search`	search	`C · · ·`	web_search, news_search
`buildkite`	devops	`· R · ·`	get_build, list_builds, list_pipelines
`circleci`	devops	`· R · ·`	get_pipeline, list_pipelines, list_workflows
`cloudflare`	cloud	`· R · ·`	list_zones, list_dns_records, list_workers
`codeberg`	development	`· R · ·`	list_repos, list_issues
`codecov`	development	`· R · ·`	get_repo_coverage, list_reports, list_repos
`cohere`	ai-ml	`· R · ·`	chat, embed, rerank
`confluence`	productivity	`· R · ·`	list_spaces, get_page, search_content
`datadog`	monitoring	`· R · ·`	query_metric, list_monitors, search_logs
`digitalocean`	cloud	`· R · ·`	list_droplets, list_apps, list_databases
`discord`	communication	`C R U D`	delete_message, edit_message, list_channels, list_guilds, list_messages, post_me
`drone`	devops	`· R · ·`	get_build, list_builds, list_repos
`elasticsearch`	database	`C R · D`	delete_doc, get_index, index_doc, list_indices, search
`fly-io`	cloud	`· R · ·`	list_apps, list_machines, get_machine
`freshdesk`	business	`C R U D`	close_ticket, create_ticket, delete_ticket, get_ticket, list_tickets, update_tic
`gemini`	ai-ml	`· R · ·`	generate_content, embed_content, list_models
`gerrit`	development	`· R · ·`	list_changes, get_change, list_reviewers
`gitea`	development	`C R · D`	close_issue, create_issue, get_repo, list_issues, list_repos
`github-actions`	devops	`· R · ·`	get_workflow_run, list_workflow_runs, list_workflows
`gitlab-ci`	devops	`· R · ·`	get_pipeline, list_jobs, list_pipelines
`gitlab`	development	`C R · ·`	list_projects, get_project, list_issues, create_issue, list_merge_requests
`google-analytics`	monitoring	`C R · ·`	run_report, list_dimensions, list_metrics
`google-drive`	storage	`C R · D`	create_file, delete_file, get_file, list_files, search_files
`grafana`	monitoring	`· R · ·`	list_dashboards, query_datasource
`groq`	ai-ml	`· R · ·`	chat_completion, list_models
`heroku`	cloud	`· R · ·`	get_app, list_apps, list_dynos
`hetzner`	cloud	`· R · ·`	list_images, list_servers, list_ssh_keys
`hubspot`	business	`C R U D`	create_contact, delete_contact, list_companies, list_contacts, list_deals, updat
`huggingface`	ai-ml	`· R · ·`	list_models, get_model, whoami
`influxdb`	database	`C R · ·`	query_flux, write_points
`jenkins`	devops	`C R · ·`	list_jobs, trigger_build, get_build
`jira`	project-management	`C R U D`	close_issue, create_issue, delete_issue, get_issue, list_projects, search_issues
`linear`	project-management	`C R · ·`	list_issues, create_issue, list_teams, list_projects
`linode`	cloud	`· R · ·`	list_domains, list_instances, list_volumes
`mailchimp`	business	`C R U D`	add_subscriber, list_audiences, list_campaigns, unsubscribe, update_subscriber
`mistral`	ai-ml	`· R · ·`	chat_completion, embed, list_models
`modal`	ai-ml	`· R · ·`	get_function_stats, list_apps
`mongodb`	database	`C R U D`	find, insert, update, delete, list_collections
`mysql`	database	`· R · ·`	query, list_tables, describe_table
`neo4j`	database	`· R · ·`	cypher, list_labels, list_relationships
`netlify`	cloud	`· R · ·`	list_sites, list_deploys, get_site
`new-relic`	monitoring	`· R · ·`	nrql, list_alerts
`notion`	productivity	`C R U D`	create_page, delete_page, get_page, query_database, search, update_page
`okta`	security	`· R · ·`	list_users, list_groups, list_apps
`openai`	ai-ml	`· R · ·`	chat_completion, embed, generate_image
`openrouter`	ai-ml	`· R · ·`	chat_completion, list_models
`pagerduty`	monitoring	`C R · ·`	list_incidents, create_incident, list_services
`prometheus`	monitoring	`· R · ·`	query, query_range, list_targets
`pulumi`	devops	`· R · ·`	get_stack, list_organizations, list_stacks
`railway`	cloud	`· R · ·`	list_projects, list_services
`redis`	database	`· R U D`	get, set, delete, keys, hget, hset
`render`	cloud	`· R · ·`	list_services, get_service, list_deploys
`replicate`	ai-ml	`C R · ·`	get_prediction, list_models, run_prediction
`s3`	storage	`· R · D`	delete_object, get_object, list_buckets, list_objects, put_object
`sendgrid`	communication	`C R · D`	delete_template, list_sent, list_templates, send_email
`sentry`	monitoring	`· R · ·`	list_issues, get_issue, list_events
`slack`	communication	`C R U D`	delete_message, edit_message, get_conversation, list_channels, list_users, post_
`snyk`	security	`· R · ·`	list_projects, list_issues
`sourcegraph`	development	`· R · ·`	get_file, search, search_symbols
`spinnaker`	devops	`C R · ·`	list_pipelines, trigger_pipeline
`sqlite`	database	`· R · ·`	query, list_tables, describe_table
`stripe`	business	`C R · D`	cancel_subscription, create_payment_intent, list_charges, list_customers, list_s
`supabase`	database	`C R · ·`	insert_into, list_tables, select_from
`tavily`	search	`· R · ·`	search, extract
`teamcity`	devops	`· R · ·`	get_build, list_builds, list_projects
`terraform`	devops	`· R · ·`	get_workspace, list_runs, list_workspaces
`together-ai`	ai-ml	`· R · ·`	chat_completion, embed, list_models
`trello`	project-management	`C R U D`	create_card, delete_card, list_boards, list_cards, update_card
`twilio`	communication	`C R · D`	delete_message, list_messages, make_call, send_sms
`vault`	security	`C R · D`	delete_secret, list_secrets, list_secrets_engines, read_secret, write_secret
`vercel`	cloud	`· R · ·`	list_deployments, list_projects, get_deployment
`wordpress`	productivity	`C R U D`	create_post, delete_post, list_pages, list_posts, update_post

Contributing

Contributions welcome! See CONTRIBUTING.md.

125 servers still planned — pick one from registry/servers/ where status: planned and implement its tools.py
More stateful mocks — wire create_* tools to local_store for cross-tool consistency
Per-tool fixtures — most servers only have default.json; add specific <method>_<path-slug>.json files for realism

Community

License

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained

ResponsivenessNo issues

Resources

GitHub Repository

Need Help?

Related Servers

Related MCP Connectors

mockmcp A
io.github.K-Cupples
Hosted MCP endpoint with realistic fake data for prototyping agents. 12 tools, no setup.
Developer Tools Testing & QA Tools
2
UnifAPIOAuth
com.unifapi
Hosted MCP server for live public-data APIs and Skills for AI agents.
Open Data Search AI & Machine Learning
2
AgentLux B
io.github.agentlux
Hosted AgentLux MCP server for marketplace, identity, creator, services, and social flows.
Agent Orchestration Autonomous Agents Social Media
7921
anotherai
dev.anotherai
MCP server for building and testing AI agents with multi-model experimentation and insights.
Agent Orchestration Autonomous Agents Testing & QA Tools
2

Related MCP Servers

datalox-gated-runtime
Agent Orchestration Testing & QA Tools Developer Tools
Oshawott324
A
license
Not graded
quality
A
maintenance
This MCP server provides a stateful, resettable, verifiable API runtime that gates every tool call, enabling agents to run long workflows against provider-shaped environments without live provider write access. It records decisions, side effects, and outcome evidence for replayable, verifiable benchmark runs.
Updated 6 days ago (2026-09-20 14:01 UTC)
Apache 2.0
scv-triage
Monitoring Observability
kc-ml2
A
license
B
quality
B
maintenance
Enables bounded, deterministic triage verdicts for server health, fleet status, and timeline investigation through a simulated execution-boundary MCP server.
Updated 2 months ago (2026-08-07 18:38 UTC)
3
5
MIT
mcp-mock
Developer Tools Testing & QA Tools
yulinlina
A
license
Not graded
quality
C
maintenance
Spin up deterministic mock MCP servers from JSON fixture files for testing MCP clients and agents.
Updated 2 months ago (2026-08-11 07:21 UTC)
MIT
mcpstub
Testing & QA Tools API Testing Developer Tools
jovial-liu
A
license
Not graded
quality
B
maintenance
Serves deterministic MCP protocol fixtures generated from tool contracts, enabling testing of MCP clients, gateways, and agent harnesses without live credentials or external services.
Updated a month ago (2026-08-23 10:27 UTC)
MIT

fastmcp-fleet

Table of contents

What is fastmcp-fleet?

Why offline-first?

Architecture

Quickstart

Docker (one container, one command — recommended)

Local (Python + uv)

Docker (all 55 services + gateway)

Server catalogue

How offline mode works

Local Store — stateful mocks

Add a new server in 5 minutes

Testing

Docker deployment

CRUD coverage

Contributing

Community

License

Maintenance

Resources

Related MCP Connectors

Related MCP Servers

datalox-gated-runtime

scv-triage

mcp-mock

mcpstub