Appwrite hit GitHub's daily trending list on November 13, gaining 78 new stars to reach 58,075 total, according to the platform's trending page (GitHub's own figure, not independently audited) S¹. The project, created in April 2019 as a backend-as-a-service, now calls itself "an MCP and agent-first, open-source platform for building and scaling apps" S¹, a repositioning that aligns it with the open-source agent tools cycling through trending boards this month.
My read: The agent-first label is doing a lot of work here. Appwrite's core is still a backend: databases, auth, storage, functions. The MCP server that backs the agent claim lives in a separate repo with 72 stars P⁵. I'd watch whether the MCP integration actually lets an agent do meaningful backend work, or whether "agent-first" is a README rebrand riding the trend. The Hugging Face presence is thin so far: one image-to-image model and no stated AI interests P⁴.
The project's README lists its capabilities as user sign-on, document storage, file hosting, serverless code execution, messaging, site hosting, live data feeds, and a web firewall S¹ (per the project's own README, not independently verified). It is available as a managed cloud service with a free tier, or as a self-hosted deployment on infrastructure the user controls S¹. The code is BSD 3-Clause licensed, with 5,869 forks and 756 open issues, according to the repository's GitHub page P².
Appwrite's agent claim rests on a 72-star side repo
The agent story hinges on Appwrite's MCP server, a separate repository created on February 25, 2025 P⁵. MCP, or Model Context Protocol, is the standard that lets AI models connect to external tools and data sources. Appwrite's MCP server, written in Python and MIT-licensed, promises to let agents operate the backend directly P⁵ — though no one outside the project has independently tested this, and no benchmarks have been released. On Appwrite Cloud, the server is included with no local installation required, according to the project's README S¹.
That MCP repo has 72 stars against the main project's 58,000-plus P⁵S¹.
The main repository's trending spike of 78 stars in a single day S¹ is modest next to the open-source AI studios and agent canvases that have dominated GitHub's daily list recently.
The backend still works without the agent pitch
A small team building an app today could use Appwrite to replace Firebase or Supabase: user authentication, a document database, file storage, and serverless functions, all behind one dashboard. The agent-first pitch adds a second access path: an AI assistant connected through MCP that can create collections, manage users, or query data without a human writing API calls. Whether that second path works reliably for production workloads is not something the README establishes.
For a solo developer already on Appwrite Cloud, the bundled MCP server means an AI coding assistant can now reach their project's collections and user tables without a separate local install — though with only 72 stars on the MCP repo, that bridge is still thinly vetted P⁵S¹.
Appwrite's Hugging Face page lists the company as verified, with one published model, appwrite/focalnet, an image-to-image model that contributor torstendittmann updated roughly 15 hours before the trending snapshot P⁴. The page lists no defined AI or ML interests, which makes the model's relationship to the backend platform unclear.
The broader pattern is familiar. Open-source projects across the agent toolchain are racing to add MCP support and rebrand around agents. Appwrite's move is the backend-as-a-service version of the same shift: take an existing infrastructure product and make it addressable by AI models.
A developer evaluating Appwrite for an agent project should check the MCP server repo directly. Its 72 stars and 7 open issues as of November 13 P⁵ are where the agent-first claim meets its first real test.
Sources: S1 — appwrite/appwrite: The open-source cloud for agents & devs. Including · P2 — appwrite/appwrite · P3 — docs/hub/security-tokens.md · P4 — appwrite (Appwrite) · P5 — appwrite/mcp
Written from 5 sourced items, 4 of them primary.
