Agenta delivers a comprehensive open-source LLMOps workspace that covers the full lifecycle of AI application development — from prompt engineering through production monitoring. The platform supports 15+ model providers including OpenAI, Anthropic, Google Gemini, Mistral, Groq, Together AI, Azure, AWS Bedrock, and self-hosted models via Ollama, enabling teams to switch between providers without code changes. The prompt playground allows side-by-side comparison of different configurations, while the evaluation system offers LLM-as-a-Judge assessment, 20+ pre-built evaluators covering semantic similarity, regex matching, and factual accuracy, plus custom Python evaluators for domain-specific requirements. Teams run evaluations through both the web UI for subject matter experts and the Evaluation SDK for programmatic CI/CD integration. The observability layer captures full trace visibility across complex agentic workflows, flagging quality issues like hallucinations and off-topic responses in real time. Human annotation workflows let domain experts review and annotate LLM outputs, feeding corrections back into the evaluation loop. The architecture supports Chain of Prompts, RAG pipelines, and multi-step agent workflows, integrating with frameworks like LangChain and LlamaIndex. Self-hosting deploys via Docker Compose with Traefik for routing, requiring only a clone, environment configuration, and a single docker compose command. On RepoCloud, deploy Agenta on a dedicated VPS with root SSH access, persistent storage for evaluation datasets and traces, and complete control over model provider credentials, all under the MIT license with no usage restrictions.

Memory OS gives Hermes Agent a permanent memory, so a new session opens with your projects, decisions, and working style already in context instead of introductions. Seven layers divide the job: workspace files carry your standing instructions into every turn, and a full-text index makes every past conversation searchable in place. Durable facts about your stack, your projects, and your patterns live in a structured store that resolves entities and scores how much each fact can be trusted, promoting what proves itself and demoting what does not. Cross-session recall pulls relevant moments from older threads, while a vector store with hybrid dense and keyword search finds the right memory even when the wording has changed. A self-curating wiki turns raw material into concept, entity, and comparison pages that are re-ingested automatically, and a Ground Truth hierarchy tells the agent to treat injected memories as authoritative instead of re-checking them. Scheduled jobs keep the system honest: hourly ingestion, weekly curation and decay scans, dead-letter reporting, and health heartbeats. It all runs beside the agent on your own machine with whichever provider you already use, from OpenRouter and Anthropic to a local Ollama. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Flowctl turns repetitive ops work, such as granting SSH access, restarting a service, or provisioning a server, into self-service forms that anyone on the team can run without holding credentials. Each workflow is a YAML or HUML file with typed inputs, validation rules, and a list of actions. Actions run in order or as a dependency graph, so independent branches execute in parallel. Every action runs in a Docker container, as a shell script, or as another flow, and scripts can target remote servers over SSH, several at once. Sensitive steps pause until a reviewer approves or rejects them. Output from one action feeds the next, files pass along as artifacts, and failed actions can be retried. Cron schedules with time zones and labels run flows unattended, while logs stream live in the browser. Email, Slack, Mattermost, and webhook notifications announce success, failure, waiting, or cancellation. Teams get isolated namespaces with user, operator, reviewer, and admin roles, OIDC single sign-on, and encrypted storage for SSH keys and secrets. A built-in flow editor means nobody has to hand-write files, and custom executors can be added as Go plugins. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

A building project usually scatters across a dozen disconnected tools, and Massing pulls the whole lifecycle into one browser workspace organized as seven rooms: Deal, Design, Planning, Schedule, Cost, Work, and Operate. Generate massing options from a zoning envelope, author the model, and produce permit-ready drawing sets with plans, sections, elevations, and schedules exported as SVG, PDF, or DXF on ARCH-D sheets, plus a MasterFormat specification manual. An IBC code pre-check flags problems before submission. The general contractor portal spans 139 modules for RFIs, submittals, change orders, G702 and G703 pay applications, a critical path schedule, safety tracking with TRIR, and COBie closeout. A development proforma handles sources and uses, XIRR, NPV, and joint venture waterfalls. A vitals strip keeps float, health, and cost per square foot in view, and a next best action banner points to the most urgent item for whichever role you pick, for example general contractor. A CAD command line speeds drafting, and an AI command bar and an MCP server connect assistants to the project. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Comparing prompts and models by eye stops working fast, so ChainForge lays every experiment out on one visual canvas. Drop a table of inputs into a Prompt Node, add a handful of models, and the app sends every combination of template variable and model, then collects the answers in an Inspect Node that groups responses by any variable. Evaluator nodes score outputs with a few lines of code, and visualization nodes plot pass rates and numeric scores as histograms or box plots. Chat Turn nodes test follow-up questions, to see if a model caves when a user pushes back. Sparkle buttons draft synthetic table rows, test questions, and starter evaluator code with a model of your choice, and an LLM scorer can judge answers against a rubric you write. Example Flows open ready-made experiments, from a no-key starter with tiny in-browser models to prompt template comparisons, sycophancy checks, gender bias audits, red-teaming runs, and 188 flows converted from OpenAI Evals. Documents, Chunking, Retrieval, and Rerank nodes cover retrieval experiments. Finished flows export as cforge files, and results export to spreadsheets or JSON for the rest of the team. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

When warehouse data never reaches the CRM, ad platform, or help desk that needs it, Multiwoven is the open-source reverse ETL platform that syncs it there on a schedule. Sources include Snowflake, BigQuery, Redshift, Databricks, PostgreSQL, ClickHouse, MariaDB, Oracle, Amazon S3, and SFTP. Models define what to move: write SQL with a 100-row preview, pick tables in a visual selector, or reuse existing dbt models. Destinations span Salesforce, HubSpot, Klaviyo, Mailchimp, Zendesk, Slack, Airtable, Google Sheets, Stripe, Facebook Custom Audiences, HTTP endpoints, and vector stores such as Pinecone, Qdrant, and Weaviate. Each sync maps model columns onto destination objects like Salesforce Account, Contact, or Opportunity, then runs as a full refresh or as an incremental sync that moves only new or changed rows, optionally tracked by a cursor field. Schedules can be manual, an interval from minutes to weeks, or a cron expression. A dashboard reports sync activity, each run records its status with successful and failed row counts, and failed runs can trigger status emails. Workspaces keep sources, models, syncs, and destinations separate for each team, while Admin, Member, and Viewer roles and email invitations control access. A connector framework lets teams add sources or destinations that are not yet supported. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

When a suspicious executable lands in your inbox, Freki gives your team a private place to pull it apart before it touches the network. Upload a sample through the web interface and Freki extracts its fingerprints at once: MD5, SHA-1, SHA-256, SHA-384, and SHA-512 hashes, plus CRC32 and SSDEEP fuzzy hashes for spotting near-duplicate variants. Static analysis then dissects Windows PE files without running them, reporting headers, sections, imported libraries, embedded strings, and capabilities identified in the code. A YARA engine sweeps every sample against your rule sets, so signatures your analysts write flag matching files the moment they arrive. Connect a VirusTotal API key and each report gains antivirus detection results from dozens of engines, placed next to your local findings. The same analysis runs through a REST API, so intake pipelines, mail gateways, and triage scripts can submit files and read verdicts automatically. Analysts discuss findings in community comments attached to each sample, building a shared memory of what a file is and why it matters. Accounts control who may submit samples and call the API, and every analyzed sample stays downloadable for re-examination or sharing with trusted peers. Docker Compose brings the whole platform up behind Nginx with its MariaDB database, keeping live malware samples on your own hardware instead of a public service. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

Middleware turns the pull request history your team already has into live DORA metrics, so engineering leads can see deployment frequency, lead time, recovery time, and change failure rate without building a reporting pipeline. Connect GitHub with a personal access token, group repositories into teams, and the dashboard starts charting how often you ship and how long changes take to reach production. Lead time is not one opaque number; it breaks into first commit to pull request, first response time, rework time, merge time, and merge to deploy, so the actual bottleneck is visible instead of guessed at. Merged pull requests count as deployments and revert pull requests count as incidents, which means recovery time and change failure rate appear even if your team has no separate incident tool wired up. You can exclude bot and dependency pull requests, define which branches count as production, and filter incidents in settings, so the metrics match how your team actually ships rather than how a generic template assumes. Bring it to a weekly engineering review to catch a review queue quietly doubling merge times, or open it after an incident to see if failure rate is trending the wrong way across a team. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.