Cloud JSON Storage for AI Agents

AI agents generate structured data constantly: tool outputs, conversation logs, scraped content, task results. Sprigr gives your agent a simple REST API to store any JSON and search it instantly. No database setup, no schema design, no backend code.

Start Free — No Credit Card

How it works

Three steps from raw agent output to searchable cloud storage.

1

Sign up and get an API key

Create a free account and generate an API key. Each key is scoped to an index, a dedicated namespace for your agent’s data.

2

Push any JSON

Send records to the Sprigr REST API with a single POST request. No schema to define, no migrations to run. If it’s valid JSON, Sprigr stores it and indexes it for search.

3

Retrieve and search

Fetch records by ID for exact lookups, or use full-text search to find records by content. Add facet filters to narrow results by any field in your JSON.

Why agents need better storage

Most AI agent frameworks leave data persistence as an exercise for the developer. The options are not great.

The current options

  • Local JSON files: lost on restart, no search, no sharing
  • Vector databases: overkill when you just need key-value + text search
  • Custom backend APIs: weeks of setup for CRUD + search
  • SQLite files: single-machine, no remote access for distributed agents
  • In-memory stores: gone when the process dies

Sprigr

  • Cloud-hosted: data persists across runs and machines
  • Schema-free JSON: store anything, change shape anytime
  • Built-in full-text search, no extra infrastructure
  • REST API: works from any language, any framework
  • Ready in 60 seconds, no provisioning, no migrations

Built for how agents actually work

Every feature designed around the reality of AI agent data patterns.

Schema-free JSON

No tables to create, no columns to define. Push a flat key-value object today, a deeply nested structure tomorrow. Sprigr indexes every field automatically.

Full-text & semantic search

Find records by content, not just IDs. Typo-tolerant keyword search plus optional hybrid semantic search that finds results by meaning, even when the exact words don’t match.

Faceted filtering

Combine text search with exact-match filters on any JSON field. Query status:complete, agent_id:agent-7, or source:web-scrape alongside free-text terms.

Multi-index support

Create separate indexes for different agents, users, or data types. Each index is isolated with no data bleed between tenants. Spin up a new index with a single API call.

REST API + MCP

Standard REST endpoints work with curl, fetch, requests, or any HTTP client. For agents built on the Model Context Protocol, Sprigr also ships an MCP server for native tool integration.

Flat pricing

Pay per records stored, not per query. No surprise bills when your agent runs a thousand searches in a loop. Free tier included for prototyping and small projects.

Store and search in two API calls

Push a JSON record and search for it. That is the entire integration. No SDK required. Any HTTP client works.

Store records
# Push JSON objects to your index
curl -X POST https://search.sprigr.com/1/indexes/agent-data/batch   -H "X-Sprigr-API-Key: YOUR_API_KEY"   -H "Content-Type: application/json"   -d '{
    "requests": [
      {
        "objectID": "task-result-001",
        "agent_id": "research-agent",
        "task": "summarize quarterly earnings",
        "status": "complete",
        "result": "Revenue grew 12% YoY to $4.2B...",
        "source": "https://example.com/earnings",
        "created_at": "2025-06-15T10:30:00Z"
      }
    ]
  }'
Search your data
# Full-text search with facet filters
curl -X POST https://search.sprigr.com/1/indexes/agent-data/query   -H "X-Sprigr-API-Key: YOUR_API_KEY"   -H "Content-Type: application/json"   -d '{
    "query": "quarterly earnings revenue",
    "filters": "agent_id:research-agent,status:complete",
    "hits_per_page": 10
  }'
Retrieve by ID
# Get a specific record by its objectID
curl https://search.sprigr.com/1/indexes/agent-data/objects/task-result-001   -H "X-Sprigr-API-Key: YOUR_API_KEY"

The same calls work from Python, JavaScript, Go, or any language with an HTTP library. Your agent stores data in one line and searches it in another.

What agents store in Sprigr

Real patterns from production AI agent deployments.

Conversation memory

Store conversation turns and summaries. Search past interactions by topic so the agent can recall relevant context from previous sessions.

Research and scraping

Persist scraped web pages, API responses, and extracted data. Full-text search lets the agent find specific facts across thousands of collected sources.

Task results and logs

Store the output of every tool call, API request, and completed task. Filter by status, agent ID, or date range to audit what happened and when.

Frequently asked questions

What kind of data can AI agents store in Sprigr?

Any valid JSON. Conversation logs, tool outputs, scraped web pages, embeddings metadata, task results, user preferences, anything your agent produces. There is no fixed schema. Each record is a JSON object with whatever fields you need, and you can change the structure at any time without migrations.

Is there a size limit per record?

Individual records can be up to 100 KB of JSON. For most AI agent use cases (conversation turns, tool call results, structured summaries) this is far more than enough. If you need to store larger payloads, split them into multiple records and link them with a shared identifier field.

How is data isolated between agents or tenants?

Every API key is scoped to a tenant, and you can further restrict keys to specific indexes using index-level ACLs. Data in one index is completely invisible to other indexes. You can create separate indexes per agent, per user, or per environment, whatever isolation model your application needs. There is no cross-index query leakage.

Can I use Sprigr without MCP?

Yes. Sprigr is a standard REST API. You can call it with curl, fetch, Python requests, or any HTTP client in any language. The MCP server is an optional integration for agents that support the Model Context Protocol. It wraps the same REST endpoints as native MCP tools.

What about structured queries beyond full-text search?

Sprigr supports faceted filtering on any field in your JSON. Combine full-text search with exact-match filters like status:complete or agent_id:agent-7. You can also retrieve any record by its unique objectID without running a search query at all.

Start storing JSON in 60 seconds

Free tier included. No credit card required. Paid plans from $49/mo.

Start Free — No Credit Card
Related MCP Search Server Claude MCP Search Sprigr Team Pricing