Pick a plugin. Make it yours.

Check tool output, track usage, change routing, or shape context. Install a maintained plugin by name—no Git or Go needed—then review its permissions and enable it in your local control plane.

Have a local model?
Try model-backed PII checks alongside your hosted agent.

No local model?
Start with usage logging or deterministic checks.

pii

Check tool output with your local model

Use an operator-bound local model for contextual checks before new tool output—including failed commands—reaches the primary provider. Your harness can ask for value-free categories and line numbers from the latest protection.

torana plugin install pii
Agent tools via MCP

Agent tools via MCP — Explain the last fresh replacement in this chat, with categories and relative output lines—not secret values.

redaction.explain_last · MCP guide →

Requested capabilities (11)
  • env.cache_get
  • env.cache_set
  • env.model_complete
  • env.plugin_config
  • env.serve_http
  • env.state_get
  • env.state_set
  • ir.cache_control.write
  • ir.tool_result_content.write
  • ir.tool_result_errors.write
  • ir.tool_results.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

usage_logger

See usage without saving prompts

Write one content-free JSONL record for each completed response, then let your harness read best-effort token totals for its verified session without exposing prompts or log contents.

torana plugin install usage_logger
Agent tools via MCP

Agent tools via MCP — Read this chat’s recorded token totals, including reported cache usage, or inspect the logging policy.

session.usage · status · MCP guide →

Requested capabilities (4)
  • env.file_append
  • env.serve_http
  • env.state_get
  • env.state_set

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

pii_guard

Block recognizable sensitive values without a model

Use deterministic checks for recognizable PII and secret shapes. It needs no model or network destination, and your harness can ask why the latest tool result was protected without receiving the sensitive value.

torana plugin install pii_guard
Agent tools via MCP

Agent tools via MCP — Explain the last fresh deterministic replacement in this chat without returning the protected value.

redaction.explain_last · MCP guide →

Requested capabilities (8)
  • env.plugin_config
  • env.serve_http
  • env.state_get
  • env.state_set
  • ir.cache_control.write
  • ir.tool_result_content.write
  • ir.tool_result_errors.write
  • ir.tool_results.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

cache_tier_selector

Choose a cache lifetime for a conversation

Choose between configured prompt-cache tiers using observed idle gaps, while keeping a live cached prefix's choice stable.

torana plugin install cache_tier_selector
Agent tools via MCP

Agent tools via MCP — Inspect configured cache-tier mode, idle-gap threshold and retention without changing cache markers.

status · MCP guide →

Requested capabilities (10)
  • env.cache_policy
  • env.host_call.torana_plugin_counter
  • env.log
  • env.now
  • env.plugin_config
  • env.serve_http
  • env.state_get
  • env.state_keys
  • env.state_set
  • ir.cache_control.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

cache_warmer

Keep one conversation's cache warm for a bounded gap

Opt a conversation into periodic refresh requests, with a deadline and break-even budget. Useful only for explicit cache markers whose lifetime refreshes on a read.

torana plugin install cache_warmer
Agent tools via MCP

Agent tools via MCP — Inspect configured opt-in count and timing without exposing conversation IDs or sending refresh requests.

status · MCP guide →

Requested capabilities (9)
  • env.background_tick
  • env.cache_policy
  • env.host_call.torana_send_request
  • env.now
  • env.plugin_config
  • env.serve_http
  • env.state_get
  • env.state_keys
  • env.state_set

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

compactor

Summarize selected historical tool results

Use an operator-bound model for eligible older tool output, only when the estimated benefit clears the cache-rewrite and summarizer-cost gate. Your harness can inspect the active limits without triggering a model call.

torana plugin install compactor
Agent tools via MCP

Agent tools via MCP — Inspect configured input limits, expected reuse and policy count. This does not run compaction or test the model.

status.get · MCP guide →

Use separately from: torana/keyword_compactor.

Requested capabilities (11)
  • env.cache_get
  • env.cache_set
  • env.emit_metric
  • env.host_call.torana_evaluate_compaction
  • env.host_call.torana_record_savings
  • env.model_complete
  • env.model_pricing
  • env.plugin_config
  • env.serve_http
  • env.shared_cache_get
  • ir.tool_results.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

decision_router

Route harder requests with explicit control

Compare configured model candidates, suggest a switch you can confirm, or apply an approved automatic escalation policy while keeping routing decisions visible.

torana plugin install decision_router
Agent tools via MCP

Agent tools via MCP — Inspect routing readiness and this session’s routing-thread summaries, without exposing prompts or configured model ladders.

status · conversation.get · MCP guide →

Requested capabilities (15)
  • env.credential_get
  • env.emit_metric
  • env.http_request
  • env.log
  • env.meta_get
  • env.meta_set
  • env.model_capabilities
  • env.plugin_config
  • env.route_request
  • env.route_request.effort
  • env.serve_http
  • env.state_get
  • env.state_keys
  • env.state_set
  • env.suggest

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

intent

Carry the reason for a tool call

Ask the model to include why it is making a tool call, strip the added field before the harness sees the streamed call, and make captured intent available to a compactor.

torana plugin install intent
Agent tools via MCP

Agent tools via MCP — Inspect the intent-field convention and history-fill mode without returning captured intent text.

status · MCP guide →

Requested capabilities (19)
  • env.cache_get
  • env.cache_set
  • env.emit_metric
  • env.log
  • env.meta_get
  • env.meta_set
  • env.plugin_config
  • env.serve_http
  • env.shared_cache_set
  • ir.cache_control.write
  • ir.messages.write.assistant
  • ir.messages.write.developer
  • ir.messages.write.other
  • ir.messages.write.system
  • ir.messages.write.tool
  • ir.messages.write.user
  • ir.stream.write
  • ir.tool_results.write
  • ir.tools.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

keyword_compactor

Trim repeatable tool output without another model

Apply explicit deterministic or keyword policies to selected tool results. Useful for long searches and listings that can be rerun, while keeping source reads exact.

torana plugin install keyword_compactor
Agent tools via MCP

Agent tools via MCP — Inspect compaction limits and policy count without exposing tool output, patterns or rerun commands.

status · MCP guide →

Use separately from: torana/compactor.

Requested capabilities (9)
  • env.cache_get
  • env.cache_set
  • env.emit_metric
  • env.host_call.torana_record_savings
  • env.model_pricing
  • env.plugin_config
  • env.serve_http
  • env.shared_cache_get
  • ir.tool_results.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

otel

Add request metrics

Emit request shape, latency, observed status classes and provider-reported token usage through Torana's metrics host.

torana plugin install otel
Agent tools via MCP

Agent tools via MCP — Inspect plugin identity and metric-emission readiness—not collector delivery or a trace browser.

status · MCP guide →

Requested capabilities (2)
  • env.emit_metric
  • env.serve_http

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

schema_translator

Adapt map-shaped tool schemas

Convert supported open-map tool parameters into key/value arrays for providers that need a constrained schema, then reverse the recorded conversion on streamed tool calls.

torana plugin install schema_translator
Requested capabilities (5)
  • env.meta_get
  • env.meta_set
  • ir.messages.write.assistant
  • ir.stream.write
  • ir.tools.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

tool_governor

Choose the tools your model sees

Restrict or replace model-visible tool definitions. Your harness can also request a reversible, user-confirmed allowance for a tool your allowlist omits; explicit denies always apply, and execution approval still belongs to the harness.

torana plugin install tool_governor
Agent tools via MCP

Agent tools via MCP — Inspect restriction counts or request a confirmed tool allowance for this session. Explicit operator denies still apply; undo stays user-controlled.

status · session.allow_tool · MCP guide →

Requested capabilities (6)
  • env.plugin_config
  • env.serve_http
  • env.state_get
  • env.state_set
  • ir.cache_control.write
  • ir.tools.write

These are the manifest’s requests, not automatic grants. Review resource bindings and the exact built digest in the control plane.

Make it your own.

Your plugin can live in your own repository. Move a harness-specific extension into a reusable plugin, then share it for a catalogue listing. For permissions, version pinning and source builds, see the installation guide. Connect MCP to let your agent discover plugin tools.

Registry API and release verification
GET https://torana.sh/registry/v1/index.json

The registry lists plugin source, manifest declarations and independently published release digests. Torana downloads release bundles from GitHub and checks them against these digests. Installation never grants permissions. How bundle verification works →