Tracker — AI Monitoring Job

A Tracker is an AI-powered monitoring job inside a Space. You tell it what to watch, and it produces scored results — quality-gated assessments you can act on.

---

What You Configure

• Name & Description — What this tracker is about
• Instruction — What to analyze (the topic and context)
• Score Meaning — The target to measure — what this tracker's score represents
• Skills — Composable capabilities (e.g., Web, Stock Analysis)
• Run Mode — Manual (on-demand) or Scheduled (automatic)

Example Instruction

Analyze market sentiment for Tesla (TSLA) stock.

Look at recent news, analyst reports, and social media.

Explain your reasoning.

Example Score Meaning

Market Sentiment

---

Custom Data Sources

Trackers can use two types of custom data:

Source Type	Description
Sources (Shared)	Named paths in your space that hold files (PDFs, CSVs, etc.); select which sources a tracker can read on the Data tab.
Linked Trackers	Link your own trackers (same account) to use their results as context.

---

Run Modes

Mode	How It Works
Manual	Trigger via Console or API
Scheduled	Run tracker automatically at your configured interval as frequent as 1 mintue

---

Tracker Run Cost

Each run costs: skill base cost × reasoning multiplier + add-ons

See skills → · See full pricing →

Create your first tracker →

---

Agent

An Agent is the AI that powers your tracker. You compose your agent's behavior by selecting skills — composable building blocks that give your agent specialized capabilities.

How It Works

1. Select skills (optional) — Choose skills when creating a tracker, or leave empty for source data only
2. Write an instruction — Tell the agent what to analyze
3. Run — The agent uses the selected skills to research, compute, and produce a scored result

Each skill brings specialized prompt instructions and domain-specific capabilities. Skills are additive — combine multiple skills for comprehensive coverage. With no skills selected, the agent operates as a general analyst using only your selected sources and linked trackers.

For the full list of available skills, search and filtering, see the Skills Catalog →.

How Skills Work

When you create or edit a tracker, you can select skills to give your agent external data access. Each skill:

1. Brings specialized capabilities — tailoring the agent's reasoning and data access for the domain
2. Enables domain-specific actions — such as searching the web, retrieving market data, or running computations
3. Adds to cost — skills are priced individually and combine additively

Skills are additive — the more you add, the broader the agent's capabilities. Use web alone for general monitoring, or combine it with financial_markets and social_media for comprehensive financial intelligence.

Skill ID Convention

Skill ID is always the snake_case version of the name — use this in the API:

"skills": ["social_media", "financial_markets"]

Reasoning Modes

Control how deeply the agent thinks:

Mode	Multiplier	Best For
Basic	×1	Quick answer — routine monitoring
Advanced	×6	Think deeper — nuanced analysis
Super	×25	Mission critical — iterative refinement
Zen	×55	Take your time — deepest research (async only)

Zen mode

zen requires async execution (sync: false). Sync requests return 400.

Using Skills

Via Console

When creating or editing a tracker, select skills from the dropdown. You can combine multiple skills.

Via API

Pass skill IDs in the agent_config.skills array:

Example: Multiple skills

curl -X POST https://api.quantinal.ai/v1/api/spaces/{space_id}/trackers \
  -H "Authorization: Bearer qapi_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Full Market Analysis",
    "description": "Comprehensive stock analysis with sentiment",
    "run_mode": "manual",
    "agent_config": {
      "instruction": "Analyze Tesla (TSLA) with financial data and social sentiment",
      "skills": ["financial_markets", "social_media", "web"],
      "reasoning_mode": "advanced"
    }
  }'

Skill Pricing

When using multiple skills, costs are additive. Base cost is then multiplied by reasoning mode and add-ons.

• Skills Catalog → — Browse all available skills with costs
• Pricing → — Full cost breakdown with examples

---

Sources (Space Data)

Sources are named paths within your space. Each source holds one or more files (PDFs, CSVs, etc.) that trackers can read during inference. Sources live on the space, not on individual trackers — upload once and attach them to any tracker (up to five per tracker).

Giving a tracker access

On the tracker's Data tab in the Console:

• Select up to five sources from the searchable list.
• Optionally add additional instructions (fs_data_hints) — free-text guidance for how the agent should use the data.

Via the API, set fs_scope_sources to an array of 1 to 5 source paths (e.g. ["/financials/", "/reports/"]) — selecting sources enables access. Omit it (or send an empty list) for trackers that don't need uploaded data. use_space_filesystem is a read-only field derived from your source selection, not an input.

Because sources are shared at the space level, multiple trackers can analyze the same dataset from different angles — for example, one source with portfolio companies and separate trackers for sentiment, risk, and fundamentals.

Source descriptions matter

Each source's description gives the AI context about what the files contain. Be specific: instead of "reports", write "Quarterly earnings reports and analyst notes for FAANG companies".

Supported Formats

Type	Formats
Documents	PDF, CSV, TXT, Markdown
Structured Data	JSON objects or arrays (via API)

Uploading Data

Create sources in Space Edit → Sources, then upload files into each source. You can also use the API:

File Upload

curl -X POST https://api.quantinal.ai/v1/api/spaces/{space_id}/files \
  -H "Authorization: Bearer qapi_..." \
  -F "file=@report.pdf" \
  -F "source=/financials/"

JSON Upload

Push structured data via API — single objects or batches of up to 1,000 rows:

curl -X POST https://api.quantinal.ai/v1/api/spaces/{space_id}/files \
  -H "Authorization: Bearer qapi_..." \
  -H "Content-Type: application/json" \
  -d '{
    "file_name": "earnings-data",
    "source": "/financials/",
    "data": [
      {"ticker": "AAPL", "eps": 1.53, "date": "2025-Q4"},
      {"ticker": "MSFT", "eps": 3.23, "date": "2025-Q4"}
    ]
  }'

Using the API

Use your space_id in the URL. The source field must match a source path (e.g. /financials/). List sources via GET /v1/api/spaces/{space_id}/sources.

Storage Pricing

Data uploads are generally free (rate limited). Storage counts against your account plan capacity. See full pricing →

Data Lifecycle

• TTL (Time To Live) — Set fs_ttl_days on the space to auto-expire data after N days (1-365). TTL applies to the whole space.
• Manual deletion — Delete files in Space Edit, or delete an entire source (removes all its files)
• GDPR compliance — Full data export and deletion available

Upload data to your space → · API Reference →