Practical overview
When it fits
Use Veldra when you want to run a self-hosted agent workspace that turns task descriptions into agents, routes requests, grounds answers in documents or coordinates specialist agents.
Who it is for
- Builders who want to define agents through natural-language task descriptions.
- Teams coordinating work among a lead agent and specialist agents.
- Users who need agents to answer questions from documents or use connected tools.
Documented uses
- Create an agent with a policy, tools, skills, knowledge base and reasoning method.
- Route a task to an existing agent or build a new one from a prompt.
- Create a team of coordinated agents for a described business task.
- Add documents to a knowledge base and ask questions with cited page or section references.
- Connect MCP tools and enable them for selected agents.
Requirements & limits
The documented deployment is self-hosted and uses a sign-in gate after a first-run administrator account is created.
- The quick-start route requires Docker; Ollama is recommended for the local-first default.
- The documented stack includes Python 3.11 or later, Vue 3, FastAPI, Postgres with pgvector, Redis and MinIO.
- The development setup additionally requires uv and Node 20.19 or later.
- MCP connector setup involves configuring credentials and enabling connectors per agent.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Veldra with the quick-start route
Bring up the local Veldra stack and complete its first-run workspace setup.
Before you start
- Docker
- Ollama for the documented local-first default
Steps
Run the command below to check prerequisites, pull local models, create
.envand start the application stack.make quickstartOpen
http://localhost:8000.Use the install wizard to name the workspace, test the provider connection and create the administrator account.
Sign in to begin building agents.
Expected outcome
The application stack is running locally and the administrator can enter the workspace.
Guide limitations
- This is the documented setup path; it does not establish that setup was independently tested.
Build and ask a document-grounded agent
Add a document to a knowledge base, build an agent for cited answers, and ask it a question.
Before you start
- A running Veldra instance
- A PDF document
- CLI authentication configured locally with
VELDRA_AUTH_ENABLED=falseorVELDRA_SERVICE_TOKEN
Steps
Add the PDF with the command below.
uv run veldra kb add ./whitepaper.pdfBuild an agent with the command below.
uv run veldra build "answer from my docs with citations"Ask a question with the command below.
uv run veldra ask "what does section 3 say about pricing?"
Expected outcome
The documented CLI flow adds a PDF, creates an agent prompt for document-grounded answers and submits a question.
Guide limitations
- The instructions do not specify a result or guarantee citation behavior for every document.
Compare options for Agent orchestration
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Veldra | Create an agent with a policy, tools, skills, knowledge base and reasoning method. | Self-hosted | The documented deployment is self-hosted and uses a sign-in gate after a first-run administrator account is created. |
| Agenite | Define typed tools with JSON Schema validation and error handling. | CLI | |
| Codefuse Muagent | Model agent workflows with intention, workflow, tool and character nodes. | Linux, macOS, Windows | The repository is public and provides a Python SDK installation command and Docker Compose setup instructions. |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- shamspias/customizable-gpt-chatbot · Official homepage title · retained in its source language